Post
1057
how is this possible
- 5 replies
AI & ML interests
None defined yet.
Recent Activity
DedeProGames
posted an update about 19 hours ago
DedeProGames
updated a
Space 4 days ago
DedeProGames
published a
Space 4 days ago
DedeProGames
posted an update 4 days ago
Post
6240
🧱 SLM Tetris Arena: can a small language model play Tetris without ever being trained on it?
I built an arena where tiny decoder-only LMs (50K–250M params) play Tetris zero-shot. There is no fine-tuning and no game data. They only use what they picked up from pre-training on text.
How it works:
- For every piece, the engine simulates each legal placement and describes the result in plain English ("clears one line, creates no new holes, keeps the stack low…").
- The model never sees the grid. It reads each description, and the arena compares log P(" good move") with log P(" bad move"). The best-rated placement is played.
- Every player gets the same piece sequence, so it's a fair race.
- There are two protocols: Guided (the rules are in the prompt) and Blind (no rules, only pre-training knowledge).
Two ways to play:
- Match: pick any models (even your own, custom architectures welcome) and watch them play side by side on retro 8-bit boards.
- Ranked: press Play and the arena picks up to 4 models at random from a curated pool of 29. Nobody chooses their opponents, so Elo can't be farmed. Matches run on the server and count even if you close the tab.
First results (~225 ranked matches):
- gpt2 (124M) leads with 1283 Elo, but SupraNeo-4M (4M) is right behind at 1239. Next come LowOnMind-5M and BananaMind-2.1-Pico (1.5M!).
- Model size barely predicts Elo (r ≈ 0.06). Survival does (r ≈ 0.9): the models that avoid holes and keep the stack low are the ones that win.
Every ranked match (seed, model commit SHAs, scores, Elo before/after) is logged in a public dataset.
▶ Play: DedeProGames/SLM-Tetris-Arena
📊 Results: DedeProGames/lm-tetris-arena-results
Want your model in the Ranked pool? Drop it in the comments!
I built an arena where tiny decoder-only LMs (50K–250M params) play Tetris zero-shot. There is no fine-tuning and no game data. They only use what they picked up from pre-training on text.
How it works:
- For every piece, the engine simulates each legal placement and describes the result in plain English ("clears one line, creates no new holes, keeps the stack low…").
- The model never sees the grid. It reads each description, and the arena compares log P(" good move") with log P(" bad move"). The best-rated placement is played.
- Every player gets the same piece sequence, so it's a fair race.
- There are two protocols: Guided (the rules are in the prompt) and Blind (no rules, only pre-training knowledge).
Two ways to play:
- Match: pick any models (even your own, custom architectures welcome) and watch them play side by side on retro 8-bit boards.
- Ranked: press Play and the arena picks up to 4 models at random from a curated pool of 29. Nobody chooses their opponents, so Elo can't be farmed. Matches run on the server and count even if you close the tab.
First results (~225 ranked matches):
- gpt2 (124M) leads with 1283 Elo, but SupraNeo-4M (4M) is right behind at 1239. Next come LowOnMind-5M and BananaMind-2.1-Pico (1.5M!).
- Model size barely predicts Elo (r ≈ 0.06). Survival does (r ≈ 0.9): the models that avoid holes and keep the stack low are the ones that win.
Every ranked match (seed, model commit SHAs, scores, Elo before/after) is logged in a public dataset.
▶ Play: DedeProGames/SLM-Tetris-Arena
📊 Results: DedeProGames/lm-tetris-arena-results
Want your model in the Ranked pool? Drop it in the comments!
- 1 reply
DedeProGames
posted an update 6 days ago
Post
2864
🧱 SLM Tetris Arena: can a small language model play Tetris without ever being trained on it?
I built an arena where tiny decoder-only LMs (50K–250M params) play Tetris zero-shot. There is no fine-tuning and no game data. They only use what they picked up from pre-training on text.
How it works:
- For every piece, the engine simulates each legal placement and describes the result in plain English ("clears one line, creates no new holes, keeps the stack low…").
- The model never sees the grid. It reads each description, and the arena compares log P(" good move") with log P(" bad move"). The best-rated placement is played.
- Every player gets the same piece sequence, so it's a fair race.
- There are two protocols: Guided (the rules are in the prompt) and Blind (no rules, only pre-training knowledge).
Two ways to play:
- Match: pick any models (even your own, custom architectures welcome) and watch them play side by side on retro 8-bit boards.
- Ranked: press Play and the arena picks up to 4 models at random from a curated pool of 29. Nobody chooses their opponents, so Elo can't be farmed. Matches run on the server and count even if you close the tab.
First results (~225 ranked matches):
- gpt2 (124M) leads with 1283 Elo, but SupraNeo-4M (4M) is right behind at 1239. Next come LowOnMind-5M and BananaMind-2.1-Pico (1.5M!).
- Model size barely predicts Elo (r ≈ 0.06). Survival does (r ≈ 0.9): the models that avoid holes and keep the stack low are the ones that win.
Every ranked match (seed, model commit SHAs, scores, Elo before/after) is logged in a public dataset.
▶ Play: DedeProGames/SLM-Tetris-Arena
📊 Results: DedeProGames/lm-tetris-arena-results
Want your model in the Ranked pool? Drop it in the comments!
I built an arena where tiny decoder-only LMs (50K–250M params) play Tetris zero-shot. There is no fine-tuning and no game data. They only use what they picked up from pre-training on text.
How it works:
- For every piece, the engine simulates each legal placement and describes the result in plain English ("clears one line, creates no new holes, keeps the stack low…").
- The model never sees the grid. It reads each description, and the arena compares log P(" good move") with log P(" bad move"). The best-rated placement is played.
- Every player gets the same piece sequence, so it's a fair race.
- There are two protocols: Guided (the rules are in the prompt) and Blind (no rules, only pre-training knowledge).
Two ways to play:
- Match: pick any models (even your own, custom architectures welcome) and watch them play side by side on retro 8-bit boards.
- Ranked: press Play and the arena picks up to 4 models at random from a curated pool of 29. Nobody chooses their opponents, so Elo can't be farmed. Matches run on the server and count even if you close the tab.
First results (~225 ranked matches):
- gpt2 (124M) leads with 1283 Elo, but SupraNeo-4M (4M) is right behind at 1239. Next come LowOnMind-5M and BananaMind-2.1-Pico (1.5M!).
- Model size barely predicts Elo (r ≈ 0.06). Survival does (r ≈ 0.9): the models that avoid holes and keep the stack low are the ones that win.
Every ranked match (seed, model commit SHAs, scores, Elo before/after) is logged in a public dataset.
▶ Play: DedeProGames/SLM-Tetris-Arena
📊 Results: DedeProGames/lm-tetris-arena-results
Want your model in the Ranked pool? Drop it in the comments!
- 1 reply
DedeProGames
posted an update 8 days ago
Post
148
🚀 Training GPT-U-20M on 2.6B tokens
DedeProGames
posted an update 9 days ago
Post
94
🚀 Possible new model in the GRM-3.2 family — GRM-3.2-Mist
A new model may be joining the GRM-3.2 family, originating from an experimental finetune currently under evaluation.
If this model passes internal testing and demonstrates strong performance, it will be released as GRM-3.2-Mist — a portable 1.7B model targeting reasoning and agentic coding tasks.
The model already exists and is currently undergoing evaluation. Follow this page for further updates:
OrionLLM
A new model may be joining the GRM-3.2 family, originating from an experimental finetune currently under evaluation.
If this model passes internal testing and demonstrates strong performance, it will be released as GRM-3.2-Mist — a portable 1.7B model targeting reasoning and agentic coding tasks.
The model already exists and is currently undergoing evaluation. Follow this page for further updates:
DedeProGames
posted an update 13 days ago
DedeProGames
posted an update 17 days ago
Post
119
@Banaxi-Tech is now a member of OrionLLM
Together, we’re going to push small models toward long-horizon agentic coding and reasoning tasks, thereby advancing AGI through efficient local LLM inference.
Thank you so much, Banaxi, for joining OrionLLM ; you are helping me make this dream a reality.
QEDOC ANANAB DNIM
Together, we’re going to push small models toward long-horizon agentic coding and reasoning tasks, thereby advancing AGI through efficient local LLM inference.
Thank you so much, Banaxi, for joining
QEDOC ANANAB DNIM
- 9 replies
DedeProGames
posted an update 18 days ago
Post
44
We are introducing BananaMind CodeQ, our family of models for coding and agentic reasoning.
Possible Confirmed sizes:
- 9B
- 2B
Please give a follow to
BananaMind if you support our work.
Possible Confirmed sizes:
- 9B
- 2B
Please give a follow to
- 41 replies
DedeProGames
posted an update 19 days ago
DedeProGames
posted an update 23 days ago
Post
126
🚀 Introducing the GRM-3.2 Family
The GRM-3.2 family is a new generation of reasoning-focused models from OrionLLM, purpose-built for long-horizon agentic tasks, extremely difficult reasoning problems, advanced coding, and local AI workflows across a wide range of hardware constraints.
GRM-3.2-Sky is the flagship model in the family: a 35B-A3B Mixture-of-Experts model built on the Ornith-1.0-35B architecture, designed for elite structured reasoning, complex multi-file coding, advanced mathematics, and sustained coherence across extended agentic workflows. It represents a substantial leap in long-horizon task capability over its predecessor, GRM-2.6-Plus.
GRM-3.2-Cliff is the mid-sized workhorse: a 9B-parameter model optimized for long-horizon agentic tasks and difficult reasoning in low-to-mid GPU environments. It delivers strong multi-step planning, debugging, and terminal-agent performance without demanding flagship-level hardware.
GRM-3.2-Turf is the lightweight edge model: a 1.2B-parameter model based on the LiquidAI/LFM2.5-1.2B-Thinking architecture, engineered for efficient on-device execution, high-fidelity instruction following, and robust tool use on mobile, embedded, and other resource-constrained hardware.
All three models are designed for users who need dependable reasoning engines that can maintain goal-directed behavior, planning quality, and task fidelity across many steps—whether on a server, a local workstation, or an edge device.
Models:
GRM-3.2-Sky: OrionLLM/GRM-3.2-Sky
GRM-3.2-Cliff: OrionLLM/GRM-3.2-Cliff
GRM-3.2-Turf: OrionLLM/GRM-3.2-Turf
Organization:
OrionLLM
The GRM-3.2 family is a new generation of reasoning-focused models from OrionLLM, purpose-built for long-horizon agentic tasks, extremely difficult reasoning problems, advanced coding, and local AI workflows across a wide range of hardware constraints.
GRM-3.2-Sky is the flagship model in the family: a 35B-A3B Mixture-of-Experts model built on the Ornith-1.0-35B architecture, designed for elite structured reasoning, complex multi-file coding, advanced mathematics, and sustained coherence across extended agentic workflows. It represents a substantial leap in long-horizon task capability over its predecessor, GRM-2.6-Plus.
GRM-3.2-Cliff is the mid-sized workhorse: a 9B-parameter model optimized for long-horizon agentic tasks and difficult reasoning in low-to-mid GPU environments. It delivers strong multi-step planning, debugging, and terminal-agent performance without demanding flagship-level hardware.
GRM-3.2-Turf is the lightweight edge model: a 1.2B-parameter model based on the LiquidAI/LFM2.5-1.2B-Thinking architecture, engineered for efficient on-device execution, high-fidelity instruction following, and robust tool use on mobile, embedded, and other resource-constrained hardware.
All three models are designed for users who need dependable reasoning engines that can maintain goal-directed behavior, planning quality, and task fidelity across many steps—whether on a server, a local workstation, or an edge device.
Models:
GRM-3.2-Sky: OrionLLM/GRM-3.2-Sky
GRM-3.2-Cliff: OrionLLM/GRM-3.2-Cliff
GRM-3.2-Turf: OrionLLM/GRM-3.2-Turf
Organization:
- 2 replies
DedeProGames
posted an update about 1 month ago
Post
129
Possible Kiyo sizes
Kiyo-230M (Kiyo-Ultra)
Kiyo-135M (Kiyo-Plus)
Kiyo-65M (Kiyo-Go)
Kiyo-15M (Kiyo-Air)
Kiyo-2M (Kiyo-Pico)
Kiyo-230M (Kiyo-Ultra)
Kiyo-135M (Kiyo-Plus)
Kiyo-65M (Kiyo-Go)
Kiyo-15M (Kiyo-Air)
Kiyo-2M (Kiyo-Pico)
LH-Tech-AI
posted an update about 2 months ago
Post
2328
Hey community and sponsors!
We are announcing the Supra3 family with four core SLM models:
- Supra3 Flash Lite: 25M parameters, ~60B pretraining tokens
- Supra3 Flash: 50M parameters, ~100B pretraining tokens
- Supra3 Pro: 75M parameters, ~150B pretraining tokens
- Supra3 Ultra: 100M parameters, ~200B pretraining tokens
For Supra3 Pro and Ultra, we search for sponsors who give us free access to compute like RTX 5090 32GB or so.
We estimate the total cost of the pro and ultra models at around $600.
For Supra3 Flash Lite and Flash, we do not need sponsors.
If anyone would apply for helping us, we would be really thankful and this person would get early access to new modele, insider information, credit and more!
Contact: here or on discord: lh_tech_ai
We are announcing the Supra3 family with four core SLM models:
- Supra3 Flash Lite: 25M parameters, ~60B pretraining tokens
- Supra3 Flash: 50M parameters, ~100B pretraining tokens
- Supra3 Pro: 75M parameters, ~150B pretraining tokens
- Supra3 Ultra: 100M parameters, ~200B pretraining tokens
For Supra3 Pro and Ultra, we search for sponsors who give us free access to compute like RTX 5090 32GB or so.
We estimate the total cost of the pro and ultra models at around $600.
For Supra3 Flash Lite and Flash, we do not need sponsors.
If anyone would apply for helping us, we would be really thankful and this person would get early access to new modele, insider information, credit and more!
Contact: here or on discord: lh_tech_ai
- 14 replies
LH-Tech-AI
posted an update 2 months ago
Post
3648
Supra2-100M is out!
Go check it out:
- https://www.reddit.com/r/LocalLLaMA/comments/1velyl9/new_models_supra2100m_base_and_instruct_go_check/
- https://huggingface.co/SupraLabs/Supra2-100M
- SupraLabs/Supra2-100M-Instruct
Give us a like and a follow!!
HAVE FUN 🤗🔥🚀
more coming soon...
Go check it out:
- https://www.reddit.com/r/LocalLLaMA/comments/1velyl9/new_models_supra2100m_base_and_instruct_go_check/
- https://huggingface.co/SupraLabs/Supra2-100M
- SupraLabs/Supra2-100M-Instruct
Give us a like and a follow!!
HAVE FUN 🤗🔥🚀
more coming soon...
- 1 reply
AxionLab-official
posted an update 2 months ago
DedeProGames
posted an update 2 months ago
Post
2467
DynamicMind-Mini-Instruct just released! A ~8.9m parameters model instruction-tuned based on DynamicMind-Mini
- 6 replies
DedeProGames
posted an update 2 months ago
Post
1566
🚀 Introducing the GRM-3.2 Family
The GRM-3.2 family is a new generation of reasoning-focused models from OrionLLM, purpose-built for long-horizon agentic tasks, extremely difficult reasoning problems, advanced coding, and local AI workflows across a wide range of hardware constraints.
GRM-3.2-Sky is the flagship model in the family: a 35B-A3B Mixture-of-Experts model built on the Ornith-1.0-35B architecture, designed for elite structured reasoning, complex multi-file coding, advanced mathematics, and sustained coherence across extended agentic workflows. It represents a substantial leap in long-horizon task capability over its predecessor, GRM-2.6-Plus.
GRM-3.2-Cliff is the mid-sized workhorse: a 9B-parameter model optimized for long-horizon agentic tasks and difficult reasoning in low-to-mid GPU environments. It delivers strong multi-step planning, debugging, and terminal-agent performance without demanding flagship-level hardware.
GRM-3.2-Turf is the lightweight edge model: a 1.2B-parameter model based on the LiquidAI/LFM2.5-1.2B-Thinking architecture, engineered for efficient on-device execution, high-fidelity instruction following, and robust tool use on mobile, embedded, and other resource-constrained hardware.
All three models are designed for users who need dependable reasoning engines that can maintain goal-directed behavior, planning quality, and task fidelity across many steps—whether on a server, a local workstation, or an edge device.
Models:
GRM-3.2-Sky: OrionLLM/GRM-3.2-Sky
GRM-3.2-Cliff: OrionLLM/GRM-3.2-Cliff
GRM-3.2-Turf: OrionLLM/GRM-3.2-Turf
Organization:
OrionLLM
The GRM-3.2 family is a new generation of reasoning-focused models from OrionLLM, purpose-built for long-horizon agentic tasks, extremely difficult reasoning problems, advanced coding, and local AI workflows across a wide range of hardware constraints.
GRM-3.2-Sky is the flagship model in the family: a 35B-A3B Mixture-of-Experts model built on the Ornith-1.0-35B architecture, designed for elite structured reasoning, complex multi-file coding, advanced mathematics, and sustained coherence across extended agentic workflows. It represents a substantial leap in long-horizon task capability over its predecessor, GRM-2.6-Plus.
GRM-3.2-Cliff is the mid-sized workhorse: a 9B-parameter model optimized for long-horizon agentic tasks and difficult reasoning in low-to-mid GPU environments. It delivers strong multi-step planning, debugging, and terminal-agent performance without demanding flagship-level hardware.
GRM-3.2-Turf is the lightweight edge model: a 1.2B-parameter model based on the LiquidAI/LFM2.5-1.2B-Thinking architecture, engineered for efficient on-device execution, high-fidelity instruction following, and robust tool use on mobile, embedded, and other resource-constrained hardware.
All three models are designed for users who need dependable reasoning engines that can maintain goal-directed behavior, planning quality, and task fidelity across many steps—whether on a server, a local workstation, or an edge device.
Models:
GRM-3.2-Sky: OrionLLM/GRM-3.2-Sky
GRM-3.2-Cliff: OrionLLM/GRM-3.2-Cliff
GRM-3.2-Turf: OrionLLM/GRM-3.2-Turf
Organization:
- 2 replies
LH-Tech-AI
posted an update 2 months ago
Post
2418
Announcing The Supra2 Family And Supra2-100M
Today, we are announcing a brand-new series of SupraLabs models: Supra2
This series will feature various models, including such as:
- 🐜 Supra2-Nano (0.4M) → The smallest Supra2 model.
- 🤏 Supra2-Small (1.4M) → The tiny model that runs everywhere.
- 💪 Supra2-Medium (25M) → Our medium class model in the Supra2 family. The powerful midsizer.
- 🔥 Supra2-Pro (100M): base, instruct, reasoning, code, math and more! → The most capable model yet! A real allrounder for all your everyday tasks.
- 🎨 Supra2-IMG → our generative text-to-image model
...and many more...
Current progress:
- Nano (0.4M) and Small (1.4M): in training; almost done. Baseline set.
- Medium (25M): coming soon...
- Pro (100M): in training; finishes in 66 hours - Monday, 3rd August 2026, 12:00AM
- IMG: coming soon...
You can support us with a like and follow if you want!
Don't miss our next release! Stay tuned...
Today, we are announcing a brand-new series of SupraLabs models: Supra2
This series will feature various models, including such as:
- 🐜 Supra2-Nano (0.4M) → The smallest Supra2 model.
- 🤏 Supra2-Small (1.4M) → The tiny model that runs everywhere.
- 💪 Supra2-Medium (25M) → Our medium class model in the Supra2 family. The powerful midsizer.
- 🔥 Supra2-Pro (100M): base, instruct, reasoning, code, math and more! → The most capable model yet! A real allrounder for all your everyday tasks.
- 🎨 Supra2-IMG → our generative text-to-image model
...and many more...
Current progress:
- Nano (0.4M) and Small (1.4M): in training; almost done. Baseline set.
- Medium (25M): coming soon...
- Pro (100M): in training; finishes in 66 hours - Monday, 3rd August 2026, 12:00AM
- IMG: coming soon...
You can support us with a like and follow if you want!
Don't miss our next release! Stay tuned...
- 30 replies
AxionLab-official
posted an update 2 months ago
Post
2379
- 3 replies