Post
4017
who's working on an NVFP4 version of Kimi-K3?
- 4 replies
AI & ML interests
Inference
Post
708
Uhh did Opus 4.8 cheat on PostTrainBench??
it found an API key in the PostTrainBench environment that allowed it to generate synthetic training data without using GPU hours, boosting the base model by 0.4913
Source: https://posttrainbench.com/traces/run.html?id=claude_non_api_max_claude-opus-4-8_10h_run1__healthbench_Qwen_Qwen3-4B-Base_17315102#tab=trace
it found an API key in the PostTrainBench environment that allowed it to generate synthetic training data without using GPU hours, boosting the base model by 0.4913
Source: https://posttrainbench.com/traces/run.html?id=claude_non_api_max_claude-opus-4-8_10h_run1__healthbench_Qwen_Qwen3-4B-Base_17315102#tab=trace
- 1 reply
Update README.md
1
#1 opened 3 months ago
by 
dorienh
julien-cΒ
submitted a
paper to Daily Papers 6 months ago
abidlabsΒ
authored 3
papers 9 months ago
Persistent Anti-Muslim Bias in Large Language Models
Paper β’ 2101.05783 β’ Published β’ 4
STG-MTL: Scalable Task Grouping for Multi-Task Learning Using Data Map
Paper β’ 2307.03374 β’ Published β’ 2
Gradio: Hassle-Free Sharing and Testing of ML Models in the Wild
Paper β’ 1906.02569 β’ Published β’ 2
Post
11556
Why I think local, open-source models will eventually win.
The most useful AI applications are moving toward multi-turn agentic behavior: systems that take hundreds or even thousands of iterative steps to complete a task, e.g. Claude Code, computer-control agents that click, type, and test repeatedly.
In these cases, the power of the model is not how smart it is per token, but in how quickly it can interact with its environment and tools across many steps. In that regime, model quality becomes secondary to latency.
An open-source model that can call tools quickly, check that the right thing was clicked, or verify that a code change actually passes tests can easily outperform a slightly βsmarterβ closed model that has to make remote API calls for every move.
Eventually, the balance tips: it becomes impractical for an agent to rely on remote inference for every micro-action. Just as no one would tolerate a keyboard that required a network request per keystroke, users wonβt accept agent workflows bottlenecked by latency. All devices will ship with local, open-source models that are βgood enoughβ and the expectation will shift toward everything running locally. Itβll happen sooner than most people think.
The most useful AI applications are moving toward multi-turn agentic behavior: systems that take hundreds or even thousands of iterative steps to complete a task, e.g. Claude Code, computer-control agents that click, type, and test repeatedly.
In these cases, the power of the model is not how smart it is per token, but in how quickly it can interact with its environment and tools across many steps. In that regime, model quality becomes secondary to latency.
An open-source model that can call tools quickly, check that the right thing was clicked, or verify that a code change actually passes tests can easily outperform a slightly βsmarterβ closed model that has to make remote API calls for every move.
Eventually, the balance tips: it becomes impractical for an agent to rely on remote inference for every micro-action. Just as no one would tolerate a keyboard that required a network request per keystroke, users wonβt accept agent workflows bottlenecked by latency. All devices will ship with local, open-source models that are βgood enoughβ and the expectation will shift toward everything running locally. Itβll happen sooner than most people think.
- 8 replies
Post
14256
deepseek-ai/DeepSeek-OCR is out! π₯ my take ‡οΈ
> pretty insane it can parse and re-render charts in HTML
> it uses CLIP and SAM features concatenated, so better grounding
> very efficient per vision tokens/performance ratio
> covers 100 languages
> pretty insane it can parse and re-render charts in HTML
> it uses CLIP and SAM features concatenated, so better grounding
> very efficient per vision tokens/performance ratio
> covers 100 languages
- 5 replies
Post
1620
What other features would you like to see on the Trackio Dashboard? ( gradio-templates/trackio-dashboard)
Post
7090
large AI labs open-sourced a ton of models last week π₯
here's few picks, find even more here merve/sep-16-releases-68d13ea4c547f02f95842f05 π€
> IBM released a new Docling model with 258M params based on Granite (A2.0) π ibm-granite/granite-docling-258M
> Xiaomi released 7B audio LM with base and instruct variants (MIT) XiaomiMiMo/mimo-audio-68cc7202692c27dae881cce0
> DecartAI released Lucy Edit, open Nano Banana π (NC) decart-ai/Lucy-Edit-Dev
> OpenGVLab released a family of agentic computer use models (3B/7B/32B) with the dataset π» OpenGVLab/scalecua-68c912cf56f7ff4c8e034003
> Meituan Longcat released thinking version of LongCat-Flash π meituan-longcat/LongCat-Flash-Thinking
here's few picks, find even more here merve/sep-16-releases-68d13ea4c547f02f95842f05 π€
> IBM released a new Docling model with 258M params based on Granite (A2.0) π ibm-granite/granite-docling-258M
> Xiaomi released 7B audio LM with base and instruct variants (MIT) XiaomiMiMo/mimo-audio-68cc7202692c27dae881cce0
> DecartAI released Lucy Edit, open Nano Banana π (NC) decart-ai/Lucy-Edit-Dev
> OpenGVLab released a family of agentic computer use models (3B/7B/32B) with the dataset π» OpenGVLab/scalecua-68c912cf56f7ff4c8e034003
> Meituan Longcat released thinking version of LongCat-Flash π meituan-longcat/LongCat-Flash-Thinking
- 2 replies
Post
3589
IBM just released small swiss army knife for the document models: granite-docling-258M on Hugging Face π₯
> not only a document converter but also can do document question answering, understand multiple languages π€―
> best part: released with Apache 2.0 license π use it with your commercial projects!
> it supports transformers, vLLM and MLX from the get-go! π€
> built on SigLIP2 & granite-165M
model: ibm-granite/granite-docling-258M
demo: ibm-granite/granite-docling-258m-demo π
> not only a document converter but also can do document question answering, understand multiple languages π€―
> best part: released with Apache 2.0 license π use it with your commercial projects!
> it supports transformers, vLLM and MLX from the get-go! π€
> built on SigLIP2 & granite-165M
model: ibm-granite/granite-docling-258M
demo: ibm-granite/granite-docling-258m-demo π
Post
1322
a ton of image/video generation models and LLMs from big labs π₯
> Meta released facebook/mobilellm-r1-68c4597b104fac45f28f448e, smol LLMs for on-device use π¬
> Tencent released tencent/SRPO, high res image generation model and tencent/POINTS-Reader, cutting edge OCR π
> ByteDance released bytedance-research/HuMo, video generation from any input β―οΈ
find more models, datasets, demos here merve/sep-11-releases-68c7dbfa26bea8cd921fa0ac
> Meta released facebook/mobilellm-r1-68c4597b104fac45f28f448e, smol LLMs for on-device use π¬
> Tencent released tencent/SRPO, high res image generation model and tencent/POINTS-Reader, cutting edge OCR π
> ByteDance released bytedance-research/HuMo, video generation from any input β―οΈ
find more models, datasets, demos here merve/sep-11-releases-68c7dbfa26bea8cd921fa0ac
Post
1105
fan-favorite vision LM Florence-2 is now officially supported in transformers π€
find all the models in
florence-community org π«‘
find all the models in
Post
1892
past week was great for open LLMs π₯ merve/sep-1-releases-68bede0e729c12597eefd050
> Google released google/embeddinggemma-300m, new embedding model with 300M params
> new update to Kimi-K2 just landed moonshotai/Kimi-K2-Instruct-0905 π
> OpenBMB released a new version to MiniCPM with 8B params openbmb/MiniCPM4.1-8B
also soooo many Qwen-Image & Kontext LoRAs dropped!
> Google released google/embeddinggemma-300m, new embedding model with 300M params
> new update to Kimi-K2 just landed moonshotai/Kimi-K2-Instruct-0905 π
> OpenBMB released a new version to MiniCPM with 8B params openbmb/MiniCPM4.1-8B
also soooo many Qwen-Image & Kontext LoRAs dropped!
Post
3786
upgrade your transformers π₯
it comes with insanely capable models like merve/sam2-66ac9deac6fca3bc5482fe30, microsoft/kosmos-2.5, and more π«‘
I built a notebook you can run with free Colab T4 to walk through the API for new models ππ»ββοΈ merve/smol-vision
fine-tuning will follow-up soon!
it comes with insanely capable models like merve/sam2-66ac9deac6fca3bc5482fe30, microsoft/kosmos-2.5, and more π«‘
I built a notebook you can run with free Colab T4 to walk through the API for new models ππ»ββοΈ merve/smol-vision
fine-tuning will follow-up soon!
Post
6344
large AI labs have dropped so many open models last week π₯ don't miss out on them
β Apple released on-device vision LMs apple/fastvlm-68ac97b9cd5cacefdd04872e & apple/mobileclip2-68ac947dcb035c54bcd20c47
β OpenGVLab released InternVL3.5, 32 new vision LMs with one based on gpt-oss! (OS) OpenGVLab/internvl35-68ac87bd52ebe953485927fb
β MSFT released a killer small TTS model (OS) microsoft/VibeVoice-1.5B
find more herehttps://huggingface.co/collections/merve/august-29-releases-68b5a3754cfb8abf59e2b486
β Apple released on-device vision LMs apple/fastvlm-68ac97b9cd5cacefdd04872e & apple/mobileclip2-68ac947dcb035c54bcd20c47
β OpenGVLab released InternVL3.5, 32 new vision LMs with one based on gpt-oss! (OS) OpenGVLab/internvl35-68ac87bd52ebe953485927fb
β MSFT released a killer small TTS model (OS) microsoft/VibeVoice-1.5B
find more herehttps://huggingface.co/collections/merve/august-29-releases-68b5a3754cfb8abf59e2b486
- 1 reply
Post
6129
first vision language model built off openai/gpt-oss-20b just dropped! π₯
InternVL3.5 comes with 32 models π€― pre-trained, fine-tuned, aligned in various sizes OpenGVLab/internvl35-68ac87bd52ebe953485927fb
comes with gpt-oss or Qwen3 for LLM part ‡οΈ
InternVL3.5 comes with 32 models π€― pre-trained, fine-tuned, aligned in various sizes OpenGVLab/internvl35-68ac87bd52ebe953485927fb
comes with gpt-oss or Qwen3 for LLM part ‡οΈ
- 1 reply
Post
3363
GPT-4.1-mini level model right in your iPhone π€―
openbmb/MiniCPM-V-4 is only 4B while surpassing GPT-4.1-mini in vision benchmarks π₯
allows commercial use as well!
openbmb/MiniCPM-V-4 is only 4B while surpassing GPT-4.1-mini in vision benchmarks π₯
allows commercial use as well!
Post
1221
we're all sleeping on this OCR model https://huggingface.co/rednote-hilab/dots.ocr π₯
dots.ocr is a new 3B model with sota performance, support for 100 languages & allowing commercial use! π€―
single e2e model to extract image, convert tables, formula, and more into markdown π
try it MohamedRashad/Dots-OCR
dots.ocr is a new 3B model with sota performance, support for 100 languages & allowing commercial use! π€―
single e2e model to extract image, convert tables, formula, and more into markdown π
try it MohamedRashad/Dots-OCR