Image-Text-to-Text
GGUF
gemma4_unified
winnow
gemma4
typed-decisions
local-inference
vision
conversational
Instructions to use EldanRing/Winnow-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use EldanRing/Winnow-12B with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf EldanRing/Winnow-12B:BF16 # Run inference directly in the terminal: llama cli -hf EldanRing/Winnow-12B:BF16
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf EldanRing/Winnow-12B:BF16 # Run inference directly in the terminal: llama cli -hf EldanRing/Winnow-12B:BF16
Use pre-built binary
# Download pre-built binary from: # https://github.com/ggerganov/llama.cpp/releases # Start a local OpenAI-compatible server with a web UI: ./llama-server -hf EldanRing/Winnow-12B:BF16 # Run inference directly in the terminal: ./llama-cli -hf EldanRing/Winnow-12B:BF16
Build from source code
git clone https://github.com/ggerganov/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-server llama-cli # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama-server -hf EldanRing/Winnow-12B:BF16 # Run inference directly in the terminal: ./build/bin/llama-cli -hf EldanRing/Winnow-12B:BF16
Use Docker
docker model run hf.co/EldanRing/Winnow-12B:BF16
- LM Studio
- Jan
- vLLM
How to use EldanRing/Winnow-12B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "EldanRing/Winnow-12B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "EldanRing/Winnow-12B", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/EldanRing/Winnow-12B:BF16
- Ollama
How to use EldanRing/Winnow-12B with Ollama:
ollama run hf.co/EldanRing/Winnow-12B:BF16
- Unsloth Desktop
- Pi
How to use EldanRing/Winnow-12B with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf EldanRing/Winnow-12B:BF16
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "EldanRing/Winnow-12B:BF16" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use EldanRing/Winnow-12B with Docker Model Runner:
docker model run hf.co/EldanRing/Winnow-12B:BF16
- Lemonade
How to use EldanRing/Winnow-12B with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull EldanRing/Winnow-12B:BF16
Run and chat with the model
lemonade run user.Winnow-12B-BF16
List all available models
lemonade list
- Hermes Agent
How to use EldanRing/Winnow-12B with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf EldanRing/Winnow-12B:BF16
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default EldanRing/Winnow-12B:BF16
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use EldanRing/Winnow-12B with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf EldanRing/Winnow-12B:BF16
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "EldanRing/Winnow-12B:BF16" \ --custom-provider-id llama-cpp \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
Download docs/benchmarks.json from EldanRing/Winnow-12B: direct link, hf CLI and curl.
- Browser
- Download file 73.6 kB
-
https://huggingface.co/EldanRing/Winnow-12B/resolve/main/docs/benchmarks.json
- Command line
-
hf download hf://EldanRing/Winnow-12B/docs/benchmarks.json
-
curl -L -o benchmarks.json https://huggingface.co/EldanRing/Winnow-12B/resolve/main/docs/benchmarks.json
73.6 kB
| { | |
| "release": "Winnow-12B", | |
| "date": "2026-09-21", | |
| "local_comparison_gpu": "NVIDIA RTX PRO 5000 Blackwell", | |
| "measurements": [ | |
| { | |
| "model": "Winnow-12B Q8", | |
| "precision": "Q8_0", | |
| "quality_percent": { | |
| "jevbench": 85.71428571428571, | |
| "kev_v9": 81.54875717017208, | |
| "kev_additions": 69.23076923076923, | |
| "typed": 70.0 | |
| }, | |
| "counts": { | |
| "jevbench": 231, | |
| "kev_v9": 1046, | |
| "kev_additions": 390, | |
| "typed": 2000 | |
| }, | |
| "calibration": { | |
| "jev_brier": 0.20571291282072968, | |
| "jev_ece": 0.0741753092521077, | |
| "kev_brier": 0.2855102865874103, | |
| "kev_ece": 0.09795411445352069, | |
| "typed_teacher_brier_non_score": 0.22546611415467793, | |
| "typed_ece": 0.15708168396777994 | |
| }, | |
| "memory_gib": { | |
| "gpu": "NVIDIA RTX PRO 5000 Blackwell", | |
| "device_baseline_mib": 2.0, | |
| "load_vram_gib": 12.82421875, | |
| "idle_vram_gib": 14.333984375, | |
| "inference_vram_gib": 14.35546875, | |
| "load_ram_pss_gib": 12.184748649597168, | |
| "idle_ram_pss_gib": 0.541377067565918, | |
| "inference_ram_pss_gib": 0.5913925170898438 | |
| }, | |
| "performance": { | |
| "performance-short": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 22.109638899564743, | |
| "p95_ms": 33.82298490032554, | |
| "range_ms": [ | |
| 21.54988795518875, | |
| 33.82298490032554 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 52.6955034583807, | |
| "p95_ms": 59.704908169806004, | |
| "range_ms": [ | |
| 51.540270913392305, | |
| 59.704908169806004 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 45.2291421195344 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 92.9789764340967, | |
| "p95_ms": 94.17951526120305, | |
| "range_ms": [ | |
| 90.98931262269616, | |
| 94.17951526120305 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 121.38708005659282, | |
| "p95_ms": 127.18812981620431, | |
| "range_ms": [ | |
| 120.11612579226494, | |
| 127.18812981620431 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 86.04095578176626 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 187.76516010984778, | |
| "p95_ms": 190.6993198208511, | |
| "range_ms": [ | |
| 185.49408204853535, | |
| 190.6993198208511 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 220.4458403866738, | |
| "p95_ms": 225.0557690858841, | |
| "range_ms": [ | |
| 218.5507151298225, | |
| 225.0557690858841 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 85.21282644042995 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 367.638946743682, | |
| "p95_ms": 372.3510168492794, | |
| "range_ms": [ | |
| 365.28318701311946, | |
| 372.3510168492794 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 399.432020727545, | |
| "p95_ms": 411.3255850970745, | |
| "range_ms": [ | |
| 398.2086409814656, | |
| 411.3255850970745 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 87.04192056754643 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 736.4177405834198, | |
| "p95_ms": 743.015272077173, | |
| "range_ms": [ | |
| 731.859543826431, | |
| 743.015272077173 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 772.390644531697, | |
| "p95_ms": 781.3249886967242, | |
| "range_ms": [ | |
| 768.2700520381331, | |
| 781.3249886967242 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 86.90719475239233 | |
| } | |
| }, | |
| "performance": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 23.32297363318503, | |
| "p95_ms": 34.777252934873104, | |
| "range_ms": [ | |
| 23.126515094190836, | |
| 34.777252934873104 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 181.3432709313929, | |
| "p95_ms": 193.60135402530432, | |
| "range_ms": [ | |
| 178.26369171962142, | |
| 193.60135402530432 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 42.87617932977263 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 100.68796202540398, | |
| "p95_ms": 101.85512574389577, | |
| "range_ms": [ | |
| 96.88412584364414, | |
| 101.85512574389577 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 256.44394662231207, | |
| "p95_ms": 266.86199894174933, | |
| "range_ms": [ | |
| 254.43977303802967, | |
| 266.86199894174933 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 79.45339084310365 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 198.03814101032913, | |
| "p95_ms": 202.5389182381332, | |
| "range_ms": [ | |
| 195.7407076843083, | |
| 202.5389182381332 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 357.35773434862494, | |
| "p95_ms": 371.45887641236186, | |
| "range_ms": [ | |
| 354.1008699685335, | |
| 371.45887641236186 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 80.79251763510284 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 391.90001506358385, | |
| "p95_ms": 398.1366469524801, | |
| "range_ms": [ | |
| 384.79119492694736, | |
| 398.1366469524801 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 549.7453005518764, | |
| "p95_ms": 553.3181391656399, | |
| "range_ms": [ | |
| 545.2003879472613, | |
| 553.3181391656399 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 81.65347989284501 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 788.8812434393913, | |
| "p95_ms": 793.8674939796329, | |
| "range_ms": [ | |
| 778.851400129497, | |
| 793.8674939796329 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 947.814577491954, | |
| "p95_ms": 950.3754759207368, | |
| "range_ms": [ | |
| 934.6124189905822, | |
| 950.3754759207368 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 81.12754680409262 | |
| } | |
| } | |
| }, | |
| "artifact": { | |
| "sha256": "b710efc4c0d048ee61eed92c5fef5ce323a4d17e7c51f9f0533cc72ae50818ea", | |
| "bytes": 12669646592 | |
| }, | |
| "serving_configuration": "Winnow llama.cpp runtime; ctx=8192; decision-context=8192; decision-parallel=4; chat-parallel=1; KV=Q8_0; selected head; optimized pipeline; memory=auto; full GPU offload; no vision projector", | |
| "source_pins": { | |
| "kev_git": "bd058057ad0aa9df3dd6d14e3542c95a5ce367b2", | |
| "laya_git": "42626c348753fbb17572a813127df2278a1ec527", | |
| "jevbench_git": "c6004e008ffba24aec091261ca1a5c02f7324702", | |
| "kev_4b": "f780095d6511e6304868ff46ca2b74ee3eba29d8", | |
| "kev_9b": "6281032426a9ca3a08374d137bf9dfd8afd78e8a", | |
| "laya": "1c5edc17a7acd8701df6fc341c0d179f1c62c982", | |
| "laya_typed": "f9ab0b228f0fc0f14d873dbc99038f135c2da1b2", | |
| "typed_decisions": "ea9306458d6e9563628369a3d1e72e362fb381d2" | |
| }, | |
| "dataset_manifest_sha256": { | |
| "kev_v9": "3c4f0be94509a3612678bfd3a30fd99a8d0ca3c47ddfe7318075d95b2fa365e4", | |
| "jevbench": "0983ad3e73a34e7c55b02b5d09ed983aeccabef0253bb31be43a2e6230d58f7b" | |
| }, | |
| "inference_build": { | |
| "server_sha256": "ab75f69c68ccc8063d44aec6ac2a24ce07e5698f8175a263973d46c029256e0f", | |
| "runtime_lock_sha256": "21e6e69d8e43e318089e77459d102f05ee2cad266999c7860e64048f69c89e84", | |
| "cuda_arch": "120", | |
| "compiler_cache_sha256": "c105481bdf985eb42f79dee77c13caf57c36e2e419acafad975c6c04b9c81adf" | |
| } | |
| }, | |
| { | |
| "model": "Winnow-12B BF16", | |
| "precision": "BF16", | |
| "quality_percent": { | |
| "jevbench": 85.28138528138528, | |
| "kev_v9": 81.45315487571702, | |
| "kev_additions": 69.23076923076923, | |
| "typed": 70.19999999999999 | |
| }, | |
| "counts": { | |
| "jevbench": 231, | |
| "kev_v9": 1046, | |
| "kev_additions": 390, | |
| "typed": 2000 | |
| }, | |
| "calibration": { | |
| "jev_brier": 0.20424101171755193, | |
| "jev_ece": 0.06547912754284907, | |
| "kev_brier": 0.28568609930480093, | |
| "kev_ece": 0.09889472147602915, | |
| "typed_teacher_brier_non_score": 0.22675176129911292, | |
| "typed_ece": 0.15511670583963688 | |
| }, | |
| "memory_gib": { | |
| "gpu": "NVIDIA RTX PRO 5000 Blackwell", | |
| "device_baseline_mib": 2.0, | |
| "load_vram_gib": 23.220703125, | |
| "idle_vram_gib": 24.771484375, | |
| "inference_vram_gib": 24.796875, | |
| "load_ram_pss_gib": 22.580784797668457, | |
| "idle_ram_pss_gib": 0.6659975051879883, | |
| "inference_ram_pss_gib": 0.7166461944580078 | |
| }, | |
| "performance": { | |
| "performance-short": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 33.53214240632951, | |
| "p95_ms": 43.883658945560455, | |
| "range_ms": [ | |
| 33.169500064104795, | |
| 43.883658945560455 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 74.17859323322773, | |
| "p95_ms": 88.81977200508118, | |
| "range_ms": [ | |
| 73.65309121087193, | |
| 88.81977200508118 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 29.822132683393367 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 107.99862560816109, | |
| "p95_ms": 110.18604878336191, | |
| "range_ms": [ | |
| 107.16228932142258, | |
| 110.18604878336191 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 147.56763633340597, | |
| "p95_ms": 153.14696403220296, | |
| "range_ms": [ | |
| 146.64038131013513, | |
| 153.14696403220296 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 74.07501674165256 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 216.6108468081802, | |
| "p95_ms": 222.67627017572522, | |
| "range_ms": [ | |
| 215.47124814242125, | |
| 222.67627017572522 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 258.43513989821076, | |
| "p95_ms": 268.3673081919551, | |
| "range_ms": [ | |
| 257.0353946648538, | |
| 268.3673081919551 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 73.86518374201641 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 427.29172739200294, | |
| "p95_ms": 429.2222601361573, | |
| "range_ms": [ | |
| 426.3705341145396, | |
| 429.2222601361573 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 467.9329860955477, | |
| "p95_ms": 476.0954240337014, | |
| "range_ms": [ | |
| 466.92013274878263, | |
| 476.0954240337014 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 74.89028677272468 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 856.0223490931094, | |
| "p95_ms": 857.3235273361206, | |
| "range_ms": [ | |
| 854.7417223453522, | |
| 857.3235273361206 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 897.3894559312612, | |
| "p95_ms": 901.5537169761956, | |
| "range_ms": [ | |
| 895.5069477669895, | |
| 901.5537169761956 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 74.764403134805 | |
| } | |
| }, | |
| "performance": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 34.854388097301126, | |
| "p95_ms": 48.964081797748804, | |
| "range_ms": [ | |
| 34.537527710199356, | |
| 48.964081797748804 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 195.44652244076133, | |
| "p95_ms": 207.5622510164976, | |
| "range_ms": [ | |
| 193.3445748873055, | |
| 207.5622510164976 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 28.69079202332727 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 112.90686530992389, | |
| "p95_ms": 116.05708813294768, | |
| "range_ms": [ | |
| 112.2279530391097, | |
| 116.05708813294768 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 273.41523696668446, | |
| "p95_ms": 277.3972353897989, | |
| "range_ms": [ | |
| 272.6574372500181, | |
| 277.3972353897989 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 70.85485880810158 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 226.20371961966157, | |
| "p95_ms": 232.85513184964657, | |
| "range_ms": [ | |
| 224.97101593762636, | |
| 232.85513184964657 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 387.9560462664813, | |
| "p95_ms": 392.6410130225122, | |
| "range_ms": [ | |
| 387.3477568849921, | |
| 392.6410130225122 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 70.73270071288998 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 446.2287644855678, | |
| "p95_ms": 449.11305233836174, | |
| "range_ms": [ | |
| 445.53177896887064, | |
| 449.11305233836174 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 607.0701309945434, | |
| "p95_ms": 612.5219059176743, | |
| "range_ms": [ | |
| 606.2590656802058, | |
| 612.5219059176743 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 71.7120960072823 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 893.4033433906734, | |
| "p95_ms": 905.7790436781943, | |
| "range_ms": [ | |
| 892.6470330916345, | |
| 905.7790436781943 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 1054.341682465747, | |
| "p95_ms": 1057.7269801869988, | |
| "range_ms": [ | |
| 1053.0990920960903, | |
| 1057.7269801869988 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 71.63617695575788 | |
| } | |
| } | |
| }, | |
| "artifact": { | |
| "sha256": "a7ec058c65b4e73da9f6066a0239b449c74d644c234a5f901aebb19e56de3a22", | |
| "bytes": 23832065792 | |
| }, | |
| "serving_configuration": "Winnow llama.cpp runtime; ctx=8192; decision-context=8192; decision-parallel=4; chat-parallel=1; KV=Q8_0; selected head; optimized pipeline; memory=auto; full GPU offload; no vision projector", | |
| "source_pins": { | |
| "kev_git": "bd058057ad0aa9df3dd6d14e3542c95a5ce367b2", | |
| "laya_git": "42626c348753fbb17572a813127df2278a1ec527", | |
| "jevbench_git": "c6004e008ffba24aec091261ca1a5c02f7324702", | |
| "kev_4b": "f780095d6511e6304868ff46ca2b74ee3eba29d8", | |
| "kev_9b": "6281032426a9ca3a08374d137bf9dfd8afd78e8a", | |
| "laya": "1c5edc17a7acd8701df6fc341c0d179f1c62c982", | |
| "laya_typed": "f9ab0b228f0fc0f14d873dbc99038f135c2da1b2", | |
| "typed_decisions": "ea9306458d6e9563628369a3d1e72e362fb381d2" | |
| }, | |
| "dataset_manifest_sha256": { | |
| "kev_v9": "3c4f0be94509a3612678bfd3a30fd99a8d0ca3c47ddfe7318075d95b2fa365e4", | |
| "jevbench": "0983ad3e73a34e7c55b02b5d09ed983aeccabef0253bb31be43a2e6230d58f7b" | |
| }, | |
| "inference_build": { | |
| "server_sha256": "ab75f69c68ccc8063d44aec6ac2a24ce07e5698f8175a263973d46c029256e0f", | |
| "runtime_lock_sha256": "21e6e69d8e43e318089e77459d102f05ee2cad266999c7860e64048f69c89e84", | |
| "cuda_arch": "120", | |
| "compiler_cache_sha256": "c105481bdf985eb42f79dee77c13caf57c36e2e419acafad975c6c04b9c81adf" | |
| } | |
| }, | |
| { | |
| "model": "Stock Gemma 4 12B IT Q8", | |
| "precision": "Q8_0", | |
| "quality_percent": { | |
| "jevbench": 83.54978354978356, | |
| "kev_v9": 78.68068833652008, | |
| "kev_additions": 66.66666666666666, | |
| "typed": 71.3 | |
| }, | |
| "counts": { | |
| "jevbench": 231, | |
| "kev_v9": 1046, | |
| "kev_additions": 390, | |
| "typed": 2000 | |
| }, | |
| "calibration": { | |
| "jev_brier": 0.27034976841267755, | |
| "jev_ece": 0.1316843789806068, | |
| "kev_brier": 0.3702287880491706, | |
| "kev_ece": 0.16323469713749836, | |
| "typed_teacher_brier_non_score": 0.316792420641541, | |
| "typed_ece": 0.24805315512618353 | |
| }, | |
| "memory_gib": { | |
| "gpu": "NVIDIA RTX PRO 5000 Blackwell", | |
| "device_baseline_mib": 2.0, | |
| "load_vram_gib": 12.82421875, | |
| "idle_vram_gib": 14.333984375, | |
| "inference_vram_gib": 14.35546875, | |
| "load_ram_pss_gib": 12.191630363464355, | |
| "idle_ram_pss_gib": 0.5481204986572266, | |
| "inference_ram_pss_gib": 0.5979843139648438 | |
| }, | |
| "performance": { | |
| "performance-short": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 22.46146695688367, | |
| "p95_ms": 30.770714860409498, | |
| "range_ms": [ | |
| 22.075897082686424, | |
| 30.770714860409498 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 52.51622456125915, | |
| "p95_ms": 60.364149045199156, | |
| "range_ms": [ | |
| 51.94238806143403, | |
| 60.364149045199156 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 44.520689673544865 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 92.13604708202183, | |
| "p95_ms": 93.77083322033286, | |
| "range_ms": [ | |
| 91.1752269603312, | |
| 93.77083322033286 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 120.54571812041104, | |
| "p95_ms": 124.4972855783999, | |
| "range_ms": [ | |
| 120.1186771504581, | |
| 124.4972855783999 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 86.82812268772719 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 185.5551926419139, | |
| "p95_ms": 188.82568273693323, | |
| "range_ms": [ | |
| 184.85999898985028, | |
| 188.82568273693323 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 217.12910407222807, | |
| "p95_ms": 222.90433011949062, | |
| "range_ms": [ | |
| 216.16374514997005, | |
| 222.90433011949062 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 86.22771355624063 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 362.9481350071728, | |
| "p95_ms": 364.25275122746825, | |
| "range_ms": [ | |
| 362.4884523451328, | |
| 364.25275122746825 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 393.8138112425804, | |
| "p95_ms": 397.9031350463629, | |
| "range_ms": [ | |
| 393.2893369346857, | |
| 397.9031350463629 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 88.16686714581849 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 727.991261985153, | |
| "p95_ms": 729.7718632034957, | |
| "range_ms": [ | |
| 726.691781077534, | |
| 729.7718632034957 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 759.8112525884062, | |
| "p95_ms": 765.421356074512, | |
| "range_ms": [ | |
| 757.9216249287128, | |
| 765.421356074512 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 87.91314311311783 | |
| } | |
| }, | |
| "performance": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 23.851238656789064, | |
| "p95_ms": 31.665181275457144, | |
| "range_ms": [ | |
| 23.5066469758749, | |
| 31.665181275457144 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 178.94592159427702, | |
| "p95_ms": 190.3981356881559, | |
| "range_ms": [ | |
| 177.06820787861943, | |
| 190.3981356881559 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 41.926543706582635 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 97.29194873943925, | |
| "p95_ms": 98.95601915195584, | |
| "range_ms": [ | |
| 96.71131893992424, | |
| 98.95601915195584 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 253.41987097635865, | |
| "p95_ms": 259.4043822027743, | |
| "range_ms": [ | |
| 251.87801895663142, | |
| 259.4043822027743 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 82.22674233224645 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 194.9084191583097, | |
| "p95_ms": 199.53563576564193, | |
| "range_ms": [ | |
| 194.04081907123327, | |
| 199.53563576564193 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 352.85464976914227, | |
| "p95_ms": 360.91970512643456, | |
| "range_ms": [ | |
| 351.35736502707005, | |
| 360.91970512643456 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 82.08983515999061 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 382.41670886054635, | |
| "p95_ms": 383.51830979809165, | |
| "range_ms": [ | |
| 381.01493706926703, | |
| 383.51830979809165 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 539.1273065470159, | |
| "p95_ms": 545.6521101295948, | |
| "range_ms": [ | |
| 538.4233240038157, | |
| 545.6521101295948 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 83.67835206612077 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 766.0538714844733, | |
| "p95_ms": 768.1685579009354, | |
| "range_ms": [ | |
| 764.6735701709986, | |
| 768.1685579009354 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 923.574757296592, | |
| "p95_ms": 928.5005992278457, | |
| "range_ms": [ | |
| 922.1836552023888, | |
| 928.5005992278457 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 83.54503825687823 | |
| } | |
| } | |
| }, | |
| "artifact": { | |
| "sha256": "f20e7ff1be28c283eeeb18fc895733791c56a5851d5cd3fe9691b7f7d12afa72", | |
| "bytes": 12669647680 | |
| }, | |
| "serving_configuration": "Winnow llama.cpp runtime; ctx=8192; decision-context=8192; decision-parallel=4; chat-parallel=1; KV=Q8_0; selected head; optimized pipeline; memory=auto; full GPU offload; no vision projector", | |
| "source_pins": { | |
| "kev_git": "bd058057ad0aa9df3dd6d14e3542c95a5ce367b2", | |
| "laya_git": "42626c348753fbb17572a813127df2278a1ec527", | |
| "jevbench_git": "c6004e008ffba24aec091261ca1a5c02f7324702", | |
| "kev_4b": "f780095d6511e6304868ff46ca2b74ee3eba29d8", | |
| "kev_9b": "6281032426a9ca3a08374d137bf9dfd8afd78e8a", | |
| "laya": "1c5edc17a7acd8701df6fc341c0d179f1c62c982", | |
| "laya_typed": "f9ab0b228f0fc0f14d873dbc99038f135c2da1b2", | |
| "typed_decisions": "ea9306458d6e9563628369a3d1e72e362fb381d2" | |
| }, | |
| "dataset_manifest_sha256": { | |
| "kev_v9": "3c4f0be94509a3612678bfd3a30fd99a8d0ca3c47ddfe7318075d95b2fa365e4", | |
| "jevbench": "0983ad3e73a34e7c55b02b5d09ed983aeccabef0253bb31be43a2e6230d58f7b" | |
| }, | |
| "inference_build": { | |
| "server_sha256": "ab75f69c68ccc8063d44aec6ac2a24ce07e5698f8175a263973d46c029256e0f", | |
| "runtime_lock_sha256": "21e6e69d8e43e318089e77459d102f05ee2cad266999c7860e64048f69c89e84", | |
| "cuda_arch": "120", | |
| "compiler_cache_sha256": "c105481bdf985eb42f79dee77c13caf57c36e2e419acafad975c6c04b9c81adf" | |
| } | |
| }, | |
| { | |
| "model": "Stock Gemma 4 12B IT BF16", | |
| "precision": "BF16", | |
| "quality_percent": { | |
| "jevbench": 83.54978354978356, | |
| "kev_v9": 78.29827915869981, | |
| "kev_additions": 65.8974358974359, | |
| "typed": 71.39999999999999 | |
| }, | |
| "counts": { | |
| "jevbench": 231, | |
| "kev_v9": 1046, | |
| "kev_additions": 390, | |
| "typed": 2000 | |
| }, | |
| "calibration": { | |
| "jev_brier": 0.26935175054650906, | |
| "jev_ece": 0.13191123135721156, | |
| "kev_brier": 0.37384468803633186, | |
| "kev_ece": 0.17031457341371484, | |
| "typed_teacher_brier_non_score": 0.3173045037688962, | |
| "typed_ece": 0.2469585567812167 | |
| }, | |
| "memory_gib": { | |
| "gpu": "NVIDIA RTX PRO 5000 Blackwell", | |
| "device_baseline_mib": 2.0, | |
| "load_vram_gib": 23.220703125, | |
| "idle_vram_gib": 24.771484375, | |
| "inference_vram_gib": 24.796875, | |
| "load_ram_pss_gib": 22.587621688842773, | |
| "idle_ram_pss_gib": 0.6732587814331055, | |
| "inference_ram_pss_gib": 0.7233238220214844 | |
| }, | |
| "performance": { | |
| "performance-short": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 34.141618525609374, | |
| "p95_ms": 41.745335794985294, | |
| "range_ms": [ | |
| 33.73613115400076, | |
| 41.745335794985294 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 75.02603204920888, | |
| "p95_ms": 89.06905213370919, | |
| "range_ms": [ | |
| 74.07003827393055, | |
| 89.06905213370919 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 29.28976548812141 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 108.13127923756838, | |
| "p95_ms": 108.66840602830052, | |
| "range_ms": [ | |
| 107.68243437632918, | |
| 108.66840602830052 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 148.21077208034694, | |
| "p95_ms": 152.73841423913836, | |
| "range_ms": [ | |
| 147.22343580797315, | |
| 152.73841423913836 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 73.9841427606133 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 216.71691467054188, | |
| "p95_ms": 223.66143111139536, | |
| "range_ms": [ | |
| 215.90717369690537, | |
| 223.66143111139536 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 258.0273279454559, | |
| "p95_ms": 268.22592318058014, | |
| "range_ms": [ | |
| 257.2539518587291, | |
| 268.22592318058014 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 73.82903187009455 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 427.47937608510256, | |
| "p95_ms": 428.38083393871784, | |
| "range_ms": [ | |
| 426.6806561499834, | |
| 428.38083393871784 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 467.7976402454078, | |
| "p95_ms": 472.7875650860369, | |
| "range_ms": [ | |
| 467.0621966943145, | |
| 472.7875650860369 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 74.85741252141587 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 855.5614438373595, | |
| "p95_ms": 856.8025063723326, | |
| "range_ms": [ | |
| 854.8735701479018, | |
| 856.8025063723326 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 896.0815852042288, | |
| "p95_ms": 901.9729560241103, | |
| "range_ms": [ | |
| 895.2084640040994, | |
| 901.9729560241103 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 74.80467996891906 | |
| } | |
| }, | |
| "performance": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 34.927206579595804, | |
| "p95_ms": 49.98836200684309, | |
| "range_ms": [ | |
| 34.57708191126585, | |
| 49.98836200684309 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 194.9117931071669, | |
| "p95_ms": 205.73526807129383, | |
| "range_ms": [ | |
| 192.28755496442318, | |
| 205.73526807129383 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 28.630975618422116 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 112.50834492966533, | |
| "p95_ms": 114.36860868707299, | |
| "range_ms": [ | |
| 111.96773778647184, | |
| 114.36860868707299 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 273.30744452774525, | |
| "p95_ms": 277.10055792704225, | |
| "range_ms": [ | |
| 272.5763861089945, | |
| 277.10055792704225 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 71.10583668261413 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 226.3854960910976, | |
| "p95_ms": 231.87217162922025, | |
| "range_ms": [ | |
| 225.70885298773646, | |
| 231.87217162922025 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 388.3433099836111, | |
| "p95_ms": 392.61183608323336, | |
| "range_ms": [ | |
| 387.42467015981674, | |
| 392.61183608323336 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 70.67590581669417 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 446.0229475516826, | |
| "p95_ms": 446.9721997156739, | |
| "range_ms": [ | |
| 444.9301161803305, | |
| 446.9721997156739 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 606.8045513238758, | |
| "p95_ms": 611.3648116588593, | |
| "range_ms": [ | |
| 606.0819658450782, | |
| 611.3648116588593 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 71.74518749686533 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 891.7435111943632, | |
| "p95_ms": 901.2053925544024, | |
| "range_ms": [ | |
| 890.0642930530012, | |
| 901.2053925544024 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 1054.1123251896352, | |
| "p95_ms": 1057.0517675951123, | |
| "range_ms": [ | |
| 1053.6570539698005, | |
| 1057.0517675951123 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 71.76951578182063 | |
| } | |
| } | |
| }, | |
| "artifact": { | |
| "sha256": "5a5eefea73350705c6753105b725a51301d653e8d6646173db29a7e8da8e6efd", | |
| "bytes": 23832066656 | |
| }, | |
| "serving_configuration": "Winnow llama.cpp runtime; ctx=8192; decision-context=8192; decision-parallel=4; chat-parallel=1; KV=Q8_0; selected head; optimized pipeline; memory=auto; full GPU offload; no vision projector", | |
| "source_pins": { | |
| "kev_git": "bd058057ad0aa9df3dd6d14e3542c95a5ce367b2", | |
| "laya_git": "42626c348753fbb17572a813127df2278a1ec527", | |
| "jevbench_git": "c6004e008ffba24aec091261ca1a5c02f7324702", | |
| "kev_4b": "f780095d6511e6304868ff46ca2b74ee3eba29d8", | |
| "kev_9b": "6281032426a9ca3a08374d137bf9dfd8afd78e8a", | |
| "laya": "1c5edc17a7acd8701df6fc341c0d179f1c62c982", | |
| "laya_typed": "f9ab0b228f0fc0f14d873dbc99038f135c2da1b2", | |
| "typed_decisions": "ea9306458d6e9563628369a3d1e72e362fb381d2" | |
| }, | |
| "dataset_manifest_sha256": { | |
| "kev_v9": "3c4f0be94509a3612678bfd3a30fd99a8d0ca3c47ddfe7318075d95b2fa365e4", | |
| "jevbench": "0983ad3e73a34e7c55b02b5d09ed983aeccabef0253bb31be43a2e6230d58f7b" | |
| }, | |
| "inference_build": { | |
| "server_sha256": "ab75f69c68ccc8063d44aec6ac2a24ce07e5698f8175a263973d46c029256e0f", | |
| "runtime_lock_sha256": "21e6e69d8e43e318089e77459d102f05ee2cad266999c7860e64048f69c89e84", | |
| "cuda_arch": "120", | |
| "compiler_cache_sha256": "c105481bdf985eb42f79dee77c13caf57c36e2e419acafad975c6c04b9c81adf" | |
| } | |
| }, | |
| { | |
| "model": "Kev-9B BF16", | |
| "precision": "BF16", | |
| "quality_percent": { | |
| "jevbench": 76.19047619047619, | |
| "kev_v9": 78.20267686424475, | |
| "kev_additions": 68.97435897435898, | |
| "typed": 71.6 | |
| }, | |
| "counts": { | |
| "jevbench": 231, | |
| "kev_v9": 1046, | |
| "kev_additions": 390, | |
| "typed": 2000 | |
| }, | |
| "calibration": { | |
| "jev_brier": 0.3540905585960517, | |
| "jev_ece": 0.16230247700094697, | |
| "kev_brier": 0.31815661474414736, | |
| "kev_ece": 0.100808307899414, | |
| "typed_teacher_brier_non_score": 0.15618699632615016, | |
| "typed_ece": 0.059057252225222456 | |
| }, | |
| "memory_gib": { | |
| "gpu": "NVIDIA RTX PRO 5000 Blackwell", | |
| "device_baseline_mib": 2.0, | |
| "load_vram_gib": 32.29296875, | |
| "idle_vram_gib": 15.23828125, | |
| "inference_vram_gib": 21.8671875, | |
| "load_ram_pss_gib": 47.3271598815918, | |
| "idle_ram_pss_gib": 1.6994895935058594, | |
| "inference_ram_pss_gib": 2.419340133666992 | |
| }, | |
| "performance": { | |
| "performance-short": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 66.0799175966531, | |
| "p95_ms": 68.83646594360471, | |
| "range_ms": [ | |
| 64.38081990927458, | |
| 68.83646594360471 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 65.56061049923301, | |
| "p95_ms": 69.4022267125547, | |
| "range_ms": [ | |
| 64.67744009569287, | |
| 69.4022267125547 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 15.133190784285258 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 72.50639772973955, | |
| "p95_ms": 77.58744899183512, | |
| "range_ms": [ | |
| 72.30309210717678, | |
| 77.58744899183512 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 72.83194921910763, | |
| "p95_ms": 76.30407717078924, | |
| "range_ms": [ | |
| 72.55592988803983, | |
| 76.30407717078924 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 110.3350911159483 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 117.60589014738798, | |
| "p95_ms": 120.17511110752821, | |
| "range_ms": [ | |
| 115.76627707108855, | |
| 120.17511110752821 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 119.11514005623758, | |
| "p95_ms": 120.19340693950653, | |
| "range_ms": [ | |
| 118.12467919662595, | |
| 120.19340693950653 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 136.04760764914255 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 229.1738234926015, | |
| "p95_ms": 231.51671793311834, | |
| "range_ms": [ | |
| 228.54166710749269, | |
| 231.51671793311834 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 231.84350552037358, | |
| "p95_ms": 234.30478060618043, | |
| "range_ms": [ | |
| 230.4078508168459, | |
| 234.30478060618043 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 139.6320029588068 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 484.2144048307091, | |
| "p95_ms": 486.98815423995256, | |
| "range_ms": [ | |
| 483.16515097394586, | |
| 486.98815423995256 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 494.19798236340284, | |
| "p95_ms": 494.99255511909723, | |
| "range_ms": [ | |
| 493.5324052348733, | |
| 494.99255511909723 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 132.17285434202574 | |
| } | |
| }, | |
| "performance": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 71.03875442408025, | |
| "p95_ms": 75.37550991401076, | |
| "range_ms": [ | |
| 70.50783606246114, | |
| 75.37550991401076 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 151.8427466508001, | |
| "p95_ms": 155.29751405119896, | |
| "range_ms": [ | |
| 151.14088403061032, | |
| 155.29751405119896 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 14.076823391782705 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 77.25602923892438, | |
| "p95_ms": 81.38543087989092, | |
| "range_ms": [ | |
| 76.64880808442831, | |
| 81.38543087989092 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 158.6230390239507, | |
| "p95_ms": 163.41657424345613, | |
| "range_ms": [ | |
| 157.5089693069458, | |
| 163.41657424345613 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 103.55178849871966 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 82.72549859248102, | |
| "p95_ms": 87.11598766967654, | |
| "range_ms": [ | |
| 82.24265603348613, | |
| 87.11598766967654 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 164.65989546850324, | |
| "p95_ms": 169.00390107184649, | |
| "range_ms": [ | |
| 164.0263837762177, | |
| 169.00390107184649 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 193.41074121316026 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 135.10010042227805, | |
| "p95_ms": 137.17515813186765, | |
| "range_ms": [ | |
| 134.41574899479747, | |
| 137.17515813186765 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 219.00241915136576, | |
| "p95_ms": 221.23930510133505, | |
| "range_ms": [ | |
| 217.83947385847569, | |
| 221.23930510133505 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 236.86140794846656 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 231.3950948882848, | |
| "p95_ms": 235.025763977319, | |
| "range_ms": [ | |
| 230.0196811556816, | |
| 235.025763977319 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 316.5164175443351, | |
| "p95_ms": 320.0176330283284, | |
| "range_ms": [ | |
| 314.9941503070295, | |
| 320.0176330283284 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 276.5832181140164 | |
| } | |
| } | |
| }, | |
| "artifact": {}, | |
| "serving_configuration": "KEV_DTYPE=bf16 KEV_MERGE=1 KEV_PREFIX_CACHE=1; unused allocator cache released after load", | |
| "source_pins": { | |
| "kev_git": "bd058057ad0aa9df3dd6d14e3542c95a5ce367b2", | |
| "laya_git": "42626c348753fbb17572a813127df2278a1ec527", | |
| "jevbench_git": "c6004e008ffba24aec091261ca1a5c02f7324702", | |
| "kev_4b": "f780095d6511e6304868ff46ca2b74ee3eba29d8", | |
| "kev_9b": "6281032426a9ca3a08374d137bf9dfd8afd78e8a", | |
| "laya": "1c5edc17a7acd8701df6fc341c0d179f1c62c982", | |
| "laya_typed": "f9ab0b228f0fc0f14d873dbc99038f135c2da1b2", | |
| "typed_decisions": "ea9306458d6e9563628369a3d1e72e362fb381d2" | |
| }, | |
| "dataset_manifest_sha256": { | |
| "kev_v9": "3c4f0be94509a3612678bfd3a30fd99a8d0ca3c47ddfe7318075d95b2fa365e4", | |
| "jevbench": "0983ad3e73a34e7c55b02b5d09ed983aeccabef0253bb31be43a2e6230d58f7b" | |
| }, | |
| "inference_build": {} | |
| }, | |
| { | |
| "model": "Kev-4B BF16", | |
| "precision": "BF16", | |
| "quality_percent": { | |
| "jevbench": 71.42857142857143, | |
| "kev_v9": 75.62141491395793, | |
| "kev_additions": 63.589743589743584, | |
| "typed": 65.75 | |
| }, | |
| "counts": { | |
| "jevbench": 231, | |
| "kev_v9": 1046, | |
| "kev_additions": 390, | |
| "typed": 2000 | |
| }, | |
| "calibration": { | |
| "jev_brier": 0.45430566532075206, | |
| "jev_ece": 0.19253563884526956, | |
| "kev_brier": 0.357785537147028, | |
| "kev_ece": 0.11677672947493024, | |
| "typed_teacher_brier_non_score": 0.22585355939972085, | |
| "typed_ece": 0.1253794714471447 | |
| }, | |
| "memory_gib": { | |
| "gpu": "NVIDIA RTX PRO 5000 Blackwell", | |
| "device_baseline_mib": 2.0, | |
| "load_vram_gib": 17.57421875, | |
| "idle_vram_gib": 8.28125, | |
| "inference_vram_gib": 14.892578125, | |
| "load_ram_pss_gib": 23.110424041748047, | |
| "idle_ram_pss_gib": 1.6300640106201172, | |
| "inference_ram_pss_gib": 2.4912405014038086 | |
| }, | |
| "performance": { | |
| "performance-short": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 65.64339064061642, | |
| "p95_ms": 67.66112381592393, | |
| "range_ms": [ | |
| 65.52932038903236, | |
| 67.66112381592393 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 65.76349004171789, | |
| "p95_ms": 67.32068909332156, | |
| "range_ms": [ | |
| 65.40474807843566, | |
| 67.32068909332156 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 15.233826136050574 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 72.70564697682858, | |
| "p95_ms": 79.85566509887576, | |
| "range_ms": [ | |
| 72.43038015440106, | |
| 79.85566509887576 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 72.54690118134022, | |
| "p95_ms": 74.15724685415626, | |
| "range_ms": [ | |
| 72.02649116516113, | |
| 74.15724685415626 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 110.03271867657838 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 77.47087301686406, | |
| "p95_ms": 78.55709921568632, | |
| "range_ms": [ | |
| 77.14755507186055, | |
| 78.55709921568632 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 77.62656570412219, | |
| "p95_ms": 78.94049910828471, | |
| "range_ms": [ | |
| 77.46943598613143, | |
| 78.94049910828471 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 206.52923320635716 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 148.88300257734954, | |
| "p95_ms": 153.7636579014361, | |
| "range_ms": [ | |
| 148.60826171934605, | |
| 153.7636579014361 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 150.5928838159889, | |
| "p95_ms": 151.1529739946127, | |
| "range_ms": [ | |
| 150.2681067213416, | |
| 151.1529739946127 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 214.9338705294781 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 312.89613409899175, | |
| "p95_ms": 314.574324991554, | |
| "range_ms": [ | |
| 312.318773008883, | |
| 314.574324991554 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 319.24319406971335, | |
| "p95_ms": 320.00197703018785, | |
| "range_ms": [ | |
| 318.71265126392245, | |
| 320.00197703018785 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 204.54071823000587 | |
| } | |
| }, | |
| "performance": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 71.24498765915632, | |
| "p95_ms": 74.90490330383182, | |
| "range_ms": [ | |
| 70.84893900901079, | |
| 74.90490330383182 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 133.07837629690766, | |
| "p95_ms": 134.00867069140077, | |
| "range_ms": [ | |
| 132.59583385661244, | |
| 134.00867069140077 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 14.036075138143156 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 78.0957960523665, | |
| "p95_ms": 79.34773107990623, | |
| "range_ms": [ | |
| 77.41492474451661, | |
| 79.34773107990623 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 139.4394184462726, | |
| "p95_ms": 142.5566808320582, | |
| "range_ms": [ | |
| 138.79002584144473, | |
| 142.5566808320582 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 102.43829251238652 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 83.62268563359976, | |
| "p95_ms": 84.26053682342172, | |
| "range_ms": [ | |
| 83.19316618144512, | |
| 84.26053682342172 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 145.1625218614936, | |
| "p95_ms": 147.97593792900443, | |
| "range_ms": [ | |
| 144.65159736573696, | |
| 147.97593792900443 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 191.33563911239858 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 102.64253127388656, | |
| "p95_ms": 103.2784916460514, | |
| "range_ms": [ | |
| 102.27458225563169, | |
| 103.2784916460514 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 164.34098314493895, | |
| "p95_ms": 165.785264223814, | |
| "range_ms": [ | |
| 163.7231931090355, | |
| 165.785264223814 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 311.7616021628762 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 179.33518323116004, | |
| "p95_ms": 180.0042991526425, | |
| "range_ms": [ | |
| 178.8237178698182, | |
| 180.0042991526425 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 240.78315845690668, | |
| "p95_ms": 242.358619812876, | |
| "range_ms": [ | |
| 240.2559290640056, | |
| 242.358619812876 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 356.8736421202139 | |
| } | |
| } | |
| }, | |
| "artifact": {}, | |
| "serving_configuration": "KEV_DTYPE=bf16 KEV_MERGE=1 KEV_PREFIX_CACHE=1; unused allocator cache released after load", | |
| "source_pins": { | |
| "kev_git": "bd058057ad0aa9df3dd6d14e3542c95a5ce367b2", | |
| "laya_git": "42626c348753fbb17572a813127df2278a1ec527", | |
| "jevbench_git": "c6004e008ffba24aec091261ca1a5c02f7324702", | |
| "kev_4b": "f780095d6511e6304868ff46ca2b74ee3eba29d8", | |
| "kev_9b": "6281032426a9ca3a08374d137bf9dfd8afd78e8a", | |
| "laya": "1c5edc17a7acd8701df6fc341c0d179f1c62c982", | |
| "laya_typed": "f9ab0b228f0fc0f14d873dbc99038f135c2da1b2", | |
| "typed_decisions": "ea9306458d6e9563628369a3d1e72e362fb381d2" | |
| }, | |
| "dataset_manifest_sha256": { | |
| "kev_v9": "3c4f0be94509a3612678bfd3a30fd99a8d0ca3c47ddfe7318075d95b2fa365e4", | |
| "jevbench": "0983ad3e73a34e7c55b02b5d09ed983aeccabef0253bb31be43a2e6230d58f7b" | |
| }, | |
| "inference_build": {} | |
| }, | |
| { | |
| "model": "Laya Typed Decisions", | |
| "precision": "float32 weights; bfloat16 autocast", | |
| "quality_percent": { | |
| "jevbench": 53.67965367965368, | |
| "kev_v9": 55.64053537284895, | |
| "kev_additions": 36.666666666666664, | |
| "typed": 76.75 | |
| }, | |
| "counts": { | |
| "jevbench": 231, | |
| "kev_v9": 1046, | |
| "kev_additions": 390, | |
| "typed": 2000 | |
| }, | |
| "calibration": { | |
| "jev_brier": 0.5087844836796537, | |
| "jev_ece": 0.07119307359307356, | |
| "kev_brier": 0.5406436657135641, | |
| "kev_ece": 0.051804615299576945, | |
| "typed_teacher_brier_non_score": 0.06574973716973499, | |
| "typed_ece": 0.2146720270839175 | |
| }, | |
| "memory_gib": { | |
| "gpu": "NVIDIA RTX PRO 5000 Blackwell", | |
| "device_baseline_mib": 2.0, | |
| "load_vram_gib": 1.822265625, | |
| "idle_vram_gib": 1.919921875, | |
| "inference_vram_gib": 4.94140625, | |
| "load_ram_pss_gib": 3.2509498596191406, | |
| "idle_ram_pss_gib": 1.0975418090820312, | |
| "inference_ram_pss_gib": 2.0477752685546875 | |
| }, | |
| "performance": { | |
| "performance-short": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 22.845126455649734, | |
| "p95_ms": 24.33396875858307, | |
| "range_ms": [ | |
| 22.740793880075216, | |
| 24.33396875858307 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 23.295121267437935, | |
| "p95_ms": 24.609151761978865, | |
| "range_ms": [ | |
| 23.227659054100513, | |
| 24.609151761978865 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 43.773012241422464 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 27.314365142956376, | |
| "p95_ms": 28.961639385670424, | |
| "range_ms": [ | |
| 27.18529710546136, | |
| 28.961639385670424 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 27.537896996363997, | |
| "p95_ms": 29.009283054620028, | |
| "range_ms": [ | |
| 27.268899139016867, | |
| 29.009283054620028 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 292.88617758934004 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 29.842145973816514, | |
| "p95_ms": 30.099600087851286, | |
| "range_ms": [ | |
| 29.773924965411425, | |
| 30.099600087851286 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 29.994444688782096, | |
| "p95_ms": 32.27252420037985, | |
| "range_ms": [ | |
| 29.866321943700314, | |
| 32.27252420037985 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 536.1544713988865 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 35.80718184821308, | |
| "p95_ms": 39.09342689439654, | |
| "range_ms": [ | |
| 35.35476792603731, | |
| 39.09342689439654 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 35.529671935364604, | |
| "p95_ms": 38.46615785732865, | |
| "range_ms": [ | |
| 35.424796398729086, | |
| 38.46615785732865 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 893.6754681127449 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 57.56832659244537, | |
| "p95_ms": 59.25153801217675, | |
| "range_ms": [ | |
| 57.05864494666457, | |
| 59.25153801217675 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 57.97937046736479, | |
| "p95_ms": 59.614480938762426, | |
| "range_ms": [ | |
| 57.77294002473354, | |
| 59.614480938762426 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 1111.7224312092242 | |
| } | |
| }, | |
| "performance": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 24.860800243914127, | |
| "p95_ms": 26.331953704357147, | |
| "range_ms": [ | |
| 24.80240212753415, | |
| 26.331953704357147 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 24.9115196056664, | |
| "p95_ms": 25.805712677538395, | |
| "range_ms": [ | |
| 24.311278015375137, | |
| 25.805712677538395 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 40.223966653881064 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 67.04071001149714, | |
| "p95_ms": 68.26437124982476, | |
| "range_ms": [ | |
| 66.94795889779925, | |
| 68.26437124982476 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 67.82145774923265, | |
| "p95_ms": 69.17453603819013, | |
| "range_ms": [ | |
| 67.67716398462653, | |
| 69.17453603819013 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 119.3304784306139 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 142.1585502102971, | |
| "p95_ms": 143.4747572056949, | |
| "range_ms": [ | |
| 141.72939117997885, | |
| 143.4747572056949 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 143.65023071877658, | |
| "p95_ms": 145.14261297881603, | |
| "range_ms": [ | |
| 143.00691802054644, | |
| 145.14261297881603 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 112.55038811475623 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 314.55507851205766, | |
| "p95_ms": 318.1276097893715, | |
| "range_ms": [ | |
| 313.771678134799, | |
| 318.1276097893715 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 317.7179682534188, | |
| "p95_ms": 319.81720635667443, | |
| "range_ms": [ | |
| 316.54817145317793, | |
| 319.81720635667443 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 101.73099144153021 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 647.3779694642872, | |
| "p95_ms": 649.9390099197626, | |
| "range_ms": [ | |
| 644.2375518381596, | |
| 649.9390099197626 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 655.2308923564851, | |
| "p95_ms": 657.3728658258915, | |
| "range_ms": [ | |
| 652.8241420164704, | |
| 657.3728658258915 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 98.86033046963391 | |
| } | |
| } | |
| }, | |
| "artifact": {}, | |
| "serving_configuration": null, | |
| "source_pins": { | |
| "kev_git": "bd058057ad0aa9df3dd6d14e3542c95a5ce367b2", | |
| "laya_git": "42626c348753fbb17572a813127df2278a1ec527", | |
| "jevbench_git": "c6004e008ffba24aec091261ca1a5c02f7324702", | |
| "kev_4b": "f780095d6511e6304868ff46ca2b74ee3eba29d8", | |
| "kev_9b": "6281032426a9ca3a08374d137bf9dfd8afd78e8a", | |
| "laya": "1c5edc17a7acd8701df6fc341c0d179f1c62c982", | |
| "laya_typed": "f9ab0b228f0fc0f14d873dbc99038f135c2da1b2", | |
| "typed_decisions": "ea9306458d6e9563628369a3d1e72e362fb381d2" | |
| }, | |
| "dataset_manifest_sha256": { | |
| "kev_v9": "3c4f0be94509a3612678bfd3a30fd99a8d0ca3c47ddfe7318075d95b2fa365e4", | |
| "jevbench": "0983ad3e73a34e7c55b02b5d09ed983aeccabef0253bb31be43a2e6230d58f7b" | |
| }, | |
| "inference_build": {} | |
| }, | |
| { | |
| "model": "Laya English", | |
| "precision": "float32 weights; bfloat16 autocast", | |
| "quality_percent": { | |
| "jevbench": 58.44155844155844, | |
| "kev_v9": 52.96367112810707, | |
| "kev_additions": 32.05128205128205, | |
| "typed": 36.1 | |
| }, | |
| "counts": { | |
| "jevbench": 231, | |
| "kev_v9": 1046, | |
| "kev_additions": 390, | |
| "typed": 2000 | |
| }, | |
| "calibration": { | |
| "jev_brier": 0.53208797, | |
| "jev_ece": 0.09038658008658014, | |
| "kev_brier": 0.5825189168117686, | |
| "kev_ece": 0.13014756076373651, | |
| "typed_teacher_brier_non_score": 0.32897452841925434, | |
| "typed_ece": 0.17523136709240314 | |
| }, | |
| "memory_gib": { | |
| "gpu": "NVIDIA RTX PRO 5000 Blackwell", | |
| "device_baseline_mib": 2.0, | |
| "load_vram_gib": 1.841796875, | |
| "idle_vram_gib": 1.919921875, | |
| "inference_vram_gib": 4.00390625, | |
| "load_ram_pss_gib": 3.2475242614746094, | |
| "idle_ram_pss_gib": 1.0939559936523438, | |
| "inference_ram_pss_gib": 2.0521316528320312 | |
| }, | |
| "performance": { | |
| "performance-short": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 23.84014753624797, | |
| "p95_ms": 24.90654308348894, | |
| "range_ms": [ | |
| 23.538042791187763, | |
| 24.90654308348894 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 24.082106538116932, | |
| "p95_ms": 24.846371728926897, | |
| "range_ms": [ | |
| 24.05818784609437, | |
| 24.846371728926897 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 41.946049137470354 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 28.08005828410387, | |
| "p95_ms": 30.12946993112564, | |
| "range_ms": [ | |
| 28.003456071019173, | |
| 30.12946993112564 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 28.088030638173223, | |
| "p95_ms": 29.704309068620205, | |
| "range_ms": [ | |
| 28.03805796429515, | |
| 29.704309068620205 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 284.8996935497389 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 30.665626749396324, | |
| "p95_ms": 33.649255987256765, | |
| "range_ms": [ | |
| 30.55564919486642, | |
| 33.649255987256765 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 30.93404253013432, | |
| "p95_ms": 32.32662100344896, | |
| "range_ms": [ | |
| 30.611341819167137, | |
| 32.32662100344896 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 521.7568233890727 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 36.21700336225331, | |
| "p95_ms": 38.06876204907894, | |
| "range_ms": [ | |
| 36.08020907267928, | |
| 38.06876204907894 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 36.47041297517717, | |
| "p95_ms": 38.297314662486315, | |
| "range_ms": [ | |
| 36.158771719783545, | |
| 38.297314662486315 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 883.5628856403833 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 57.41862556897104, | |
| "p95_ms": 58.74502472579479, | |
| "range_ms": [ | |
| 57.02610267326236, | |
| 58.74502472579479 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 57.85714997909963, | |
| "p95_ms": 58.998359367251396, | |
| "range_ms": [ | |
| 57.67525406554341, | |
| 58.998359367251396 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 1114.6208981112486 | |
| } | |
| }, | |
| "performance": { | |
| "1": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 25.882226647809148, | |
| "p95_ms": 26.91890113055706, | |
| "range_ms": [ | |
| 25.42299311608076, | |
| 26.91890113055706 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 25.474075926467776, | |
| "p95_ms": 25.56223887950182, | |
| "range_ms": [ | |
| 25.444936007261276, | |
| 25.56223887950182 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 38.63655216405606 | |
| }, | |
| "8": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 45.27222109027207, | |
| "p95_ms": 47.11010167375207, | |
| "range_ms": [ | |
| 45.21714290603995, | |
| 47.11010167375207 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 45.3107540961355, | |
| "p95_ms": 46.22122226282954, | |
| "range_ms": [ | |
| 45.23011017590761, | |
| 46.22122226282954 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 176.7088030438827 | |
| }, | |
| "16": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 89.15169816464186, | |
| "p95_ms": 90.1275728829205, | |
| "range_ms": [ | |
| 88.40252412483096, | |
| 90.1275728829205 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 88.64646148867905, | |
| "p95_ms": 90.57755582034588, | |
| "range_ms": [ | |
| 88.46736513078213, | |
| 90.57755582034588 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 179.4693800498542 | |
| }, | |
| "32": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 191.60822709091008, | |
| "p95_ms": 194.2352126352489, | |
| "range_ms": [ | |
| 191.27637008205056, | |
| 194.2352126352489 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 191.81448104791343, | |
| "p95_ms": 193.55412013828754, | |
| "range_ms": [ | |
| 191.48092670366168, | |
| 193.55412013828754 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 167.00744266485665 | |
| }, | |
| "64": { | |
| "warm": { | |
| "count": 10, | |
| "median_ms": 412.5161219853908, | |
| "p95_ms": 416.34345101192594, | |
| "range_ms": [ | |
| 411.5443374030292, | |
| 416.34345101192594 | |
| ] | |
| }, | |
| "cold": { | |
| "count": 10, | |
| "median_ms": 415.1305085979402, | |
| "p95_ms": 416.15893226116896, | |
| "range_ms": [ | |
| 411.9292129762471, | |
| 416.15893226116896 | |
| ] | |
| }, | |
| "warm_median_decisions_per_second": 155.1454515085996 | |
| } | |
| } | |
| }, | |
| "artifact": {}, | |
| "serving_configuration": null, | |
| "source_pins": { | |
| "kev_git": "bd058057ad0aa9df3dd6d14e3542c95a5ce367b2", | |
| "laya_git": "42626c348753fbb17572a813127df2278a1ec527", | |
| "jevbench_git": "c6004e008ffba24aec091261ca1a5c02f7324702", | |
| "kev_4b": "f780095d6511e6304868ff46ca2b74ee3eba29d8", | |
| "kev_9b": "6281032426a9ca3a08374d137bf9dfd8afd78e8a", | |
| "laya": "1c5edc17a7acd8701df6fc341c0d179f1c62c982", | |
| "laya_typed": "f9ab0b228f0fc0f14d873dbc99038f135c2da1b2", | |
| "typed_decisions": "ea9306458d6e9563628369a3d1e72e362fb381d2" | |
| }, | |
| "dataset_manifest_sha256": { | |
| "kev_v9": "3c4f0be94509a3612678bfd3a30fd99a8d0ca3c47ddfe7318075d95b2fa365e4", | |
| "jevbench": "0983ad3e73a34e7c55b02b5d09ed983aeccabef0253bb31be43a2e6230d58f7b" | |
| }, | |
| "inference_build": {} | |
| } | |
| ], | |
| "hardware_5070ti": { | |
| "gpu": "RTX 5070 Ti 16 GB", | |
| "prefix_positions": 65022, | |
| "image_positions": 1024, | |
| "decision_branches": 4, | |
| "chat_slots": 1, | |
| "kv": "Q8_0", | |
| "effective_memory_policy": "exclusive", | |
| "decision_cold_s": 25.001564278034493, | |
| "decision_warm_median_ms": 143.02524807862937, | |
| "decision_warm_samples": 3, | |
| "retrieval_passed": true, | |
| "short_generation_tps_median": 55.46540467494125, | |
| "short_generation_samples": 3, | |
| "short_output_tokens_per_sample": 512, | |
| "long_prompt_tokens": 62435, | |
| "long_output_tokens": 512, | |
| "long_generation_tps": 46.93347166614375, | |
| "long_prefill_tps": 2893.8818373839104, | |
| "long_ttft_s": 21.747452786192298, | |
| "long_total_s": 32.63961890107021, | |
| "smoke_peak_device_vram_gib": 15.0087890625, | |
| "smoke_peak_ram_pss_gib_including_startup": 12.247476577758789, | |
| "smoke_device_baseline_mib": 450.0 | |
| }, | |
| "jev": { | |
| "label": "Jev 1.13 (OpenRouter)", | |
| "model_revisions": [ | |
| "typesafe/jev-1.13-20260917" | |
| ], | |
| "quality": { | |
| "jevbench": 85.71428571428571, | |
| "kev_v9": 86.9980879541109, | |
| "kev_additions": 85.64102564102564, | |
| "typed": 73.8 | |
| }, | |
| "counts": { | |
| "jevbench": 231, | |
| "jevbench_correct": 198, | |
| "kev_v9": 1046, | |
| "kev_additions": 390, | |
| "typed_cases": 400, | |
| "typed": 2000 | |
| }, | |
| "calibration": { | |
| "jev_brier": 0.17946493506493508, | |
| "jev_ece": 0.04740259740259756, | |
| "kev_brier": 0.19142599481108083, | |
| "kev_ece": 0.030807501400235617, | |
| "typed_teacher_brier_non_score": 0.09788350992788927, | |
| "typed_ece": 0.03695434343434338 | |
| }, | |
| "coverage": { | |
| "requests": 1895, | |
| "jevbench": 231, | |
| "kev_v9": 1264, | |
| "typed_cases": 400, | |
| "typed_decisions": 2000, | |
| "unique_api_requests": 1840, | |
| "failed_requests": 0, | |
| "jevbench_schema_validity": 1.0, | |
| "contract_warnings": 0 | |
| }, | |
| "cost_usd": 0.04605409200000002, | |
| "input_tokens": 1096526, | |
| "output_tokens": 114470, | |
| "pins": { | |
| "kev_git": "bd058057ad0aa9df3dd6d14e3542c95a5ce367b2", | |
| "laya_git": "42626c348753fbb17572a813127df2278a1ec527", | |
| "jevbench_git": "c6004e008ffba24aec091261ca1a5c02f7324702", | |
| "kev_4b": "f780095d6511e6304868ff46ca2b74ee3eba29d8", | |
| "kev_9b": "6281032426a9ca3a08374d137bf9dfd8afd78e8a", | |
| "laya": "1c5edc17a7acd8701df6fc341c0d179f1c62c982", | |
| "laya_typed": "f9ab0b228f0fc0f14d873dbc99038f135c2da1b2", | |
| "typed_decisions": "ea9306458d6e9563628369a3d1e72e362fb381d2" | |
| }, | |
| "collection_date": "2026-09-21", | |
| "latency_scope": "OpenRouter round trip; not comparable to pod-local latency", | |
| "gpu_memory_gib": null, | |
| "system_memory_gib": null, | |
| "metric_scope": "JevBench public-subset accuracy; Kev-v9 knowable clean accuracy; typed teacher agreement.", | |
| "deduplication": "Identical payloads reuse one frozen response; all evaluation records remain scored." | |
| }, | |
| "fine_tuning_overlap_audit": { | |
| "by_suite": { | |
| "jevbench": { | |
| "evaluated_records": 231, | |
| "exact_normalized_input": 0, | |
| "exact_normalized_state": 0, | |
| "high_containment_state_candidates": 0 | |
| }, | |
| "kev-v9": { | |
| "evaluated_records": 1264, | |
| "exact_normalized_input": 0, | |
| "exact_normalized_state": 0, | |
| "high_containment_state_candidates": 0 | |
| }, | |
| "typed-decisions": { | |
| "evaluated_records": 400, | |
| "exact_normalized_input": 0, | |
| "exact_normalized_state": 0, | |
| "high_containment_state_candidates": 0 | |
| } | |
| }, | |
| "method": "Casefolded alphanumeric exact state/input matching; question IDs and label fields excluded; typed cases split into questions; 5-word state containment >=85% with minimum20 shingles.", | |
| "scope": "Checksum-verified private fine-tuning inputs. Does not audit Gemma pretraining or exclude conceptual/source-family overlap or development exposure. The training data and training pipeline remain private and are not distributed." | |
| } | |
| } | |