Post
5667
how is this possible
- 13 replies
Orion LLM Labs
community
AI & ML interests
Advancing AGI through efficient local LLM inference.
Recent Activity
DedeProGames
posted an update 3 days ago
Banaxi-Tech
posted an update 4 days ago
DedeProGames
posted an update 5 days ago
Post
6256
🧱 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
Banaxi-Tech
posted an update 8 days ago
Post
7592
We're releasing a MAJOR update to the BananaAll SLM Super App.
If you want to use a custom architecture, previously you had to go trough reviewing the code yourself, now add an Openrouter API key and review it with GPT 6 Luna in one button. A review cost be half a cent so anyone can try it. This is one of the main features.
Now ROCm, AMD and Windows, Mac support.
Colab and Molab support.
Detailed list of features:
Get improved Windows Python detection and support paths for compatible AMD ROCm, Intel XPU, and Apple MPS setups.
Choose local training or export a self-contained Python script for Colab or Molab. Notebook runs produce a downloadable model ZIP.
Start pretraining with an existing model’s tokenizer, or train a new one from your datasets.
Try experimental 1.58-bit Ternary fake-quantized training on NVIDIA GPUs.
Watch live tokens per second. Model compilation is on by default and falls back automatically if it fails.
Build custom architectures with separate configuration and modeling files, then review the training code manually or with optional OpenRouter AI Review.
Install from source with the new coding-agent instructions.
This release also fixes inflated loss reporting for custom models.
And for those users who didn't want to try it out just because installation would be so hard, it isnt now.
Go to any coding agent (Pi, Claude Code, Codex, OpenCode, basically all work), and just paste "Install BananaAll for me. Fetch and follow https://raw.githubusercontent.com/BananaMind/BananaAll/main/agent_install.txt."
That's it.
Check it out at https://github.com/BananaMind/BananaAll/
Also on SAICR, we're currently training a new major model (NACR v2) and ACR 1.0 is in the finishing.
If you want to use a custom architecture, previously you had to go trough reviewing the code yourself, now add an Openrouter API key and review it with GPT 6 Luna in one button. A review cost be half a cent so anyone can try it. This is one of the main features.
Now ROCm, AMD and Windows, Mac support.
Colab and Molab support.
Detailed list of features:
Get improved Windows Python detection and support paths for compatible AMD ROCm, Intel XPU, and Apple MPS setups.
Choose local training or export a self-contained Python script for Colab or Molab. Notebook runs produce a downloadable model ZIP.
Start pretraining with an existing model’s tokenizer, or train a new one from your datasets.
Try experimental 1.58-bit Ternary fake-quantized training on NVIDIA GPUs.
Watch live tokens per second. Model compilation is on by default and falls back automatically if it fails.
Build custom architectures with separate configuration and modeling files, then review the training code manually or with optional OpenRouter AI Review.
Install from source with the new coding-agent instructions.
This release also fixes inflated loss reporting for custom models.
And for those users who didn't want to try it out just because installation would be so hard, it isnt now.
Go to any coding agent (Pi, Claude Code, Codex, OpenCode, basically all work), and just paste "Install BananaAll for me. Fetch and follow https://raw.githubusercontent.com/BananaMind/BananaAll/main/agent_install.txt."
That's it.
Check it out at https://github.com/BananaMind/BananaAll/
Also on SAICR, we're currently training a new major model (NACR v2) and ACR 1.0 is in the finishing.
- 11 replies
DedeProGames
posted an update 8 days ago
Post
2873
🧱 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 9 days ago
Post
150
🚀 Training GPT-U-20M on 2.6B tokens
Banaxi-Tech
posted an update 10 days ago
Post
4862
We're excited to release BananaAll, our SLM Super App.
It allows you to do EVERYTHING you need to do to trains SLMs in a single app, no terminal, no 30 chrome tabs.
The train tab allows you to train models, select datasets from presets, and use other ones with auto mapping, model size slider, it automatically generates a training script for you.
Then after you've trained the model or want to compare it to competitors, the evaluation tab, run ARC EASY, ARC Challenge, Hellaswag, PIQA, Arithmark 3, BananaMind Base Bench and more! Simple Results screen.
And lastly the inference tab, run your trained models or others.
Normally you would need seperate apps or scripts for that, but the BananaAll Super App lets you do all of that in a single app.
We also trained a small 2.5M parameter model on 200M tokens of Fineweb edu, The results: BananaMind Base Bench 854 and 53% on PIQA. On only 200M tokens.
Check it out at https://github.com/BananaMind/BananaAll.
It allows you to do EVERYTHING you need to do to trains SLMs in a single app, no terminal, no 30 chrome tabs.
The train tab allows you to train models, select datasets from presets, and use other ones with auto mapping, model size slider, it automatically generates a training script for you.
Then after you've trained the model or want to compare it to competitors, the evaluation tab, run ARC EASY, ARC Challenge, Hellaswag, PIQA, Arithmark 3, BananaMind Base Bench and more! Simple Results screen.
And lastly the inference tab, run your trained models or others.
Normally you would need seperate apps or scripts for that, but the BananaAll Super App lets you do all of that in a single app.
We also trained a small 2.5M parameter model on 200M tokens of Fineweb edu, The results: BananaMind Base Bench 854 and 53% on PIQA. On only 200M tokens.
Check it out at https://github.com/BananaMind/BananaAll.
- 25 replies
DedeProGames
posted an update 10 days ago
Post
95
🚀 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:
Banaxi-Tech
posted an update 11 days ago
Post
1899
hi everyone
we have released nacr
its not just any model, its nacr
we have 6 more features and this model only uses 20% of its total capacity!
check it out at saicr/nacr
we're currently working on expanding access as we do more research but right now you have to use our gated access form
follow
saicr if you're interested
if you want to join saicr, first read the entire nacr readme, then press the join button.
we have released nacr
its not just any model, its nacr
we have 6 more features and this model only uses 20% of its total capacity!
check it out at saicr/nacr
we're currently working on expanding access as we do more research but right now you have to use our gated access form
follow
if you want to join saicr, first read the entire nacr readme, then press the join button.
- 14 replies
Banaxi-Tech
posted an update 12 days ago
DedeProGames
updated 3
models 12 days ago
Banaxi-Tech
posted an update 13 days ago
Post
101
This day is Sol nice.
This is a good model, This is my assessment of the model
5
#8 opened about 1 month ago
by 
funnygeeker
Banaxi-Tech
posted an update 14 days ago
Post
103
We have some updates to @BananaMindBot 🍌
It can now train models, ask it to train a model, and i will train it for you.
It now can also merge PRs And like models.
It can now train models, ask it to train a model, and i will train it for you.
It now can also merge PRs And like models.
- 28 replies
DedeProGames
posted an update 14 days ago
Banaxi-Tech
posted an update 15 days ago
Post
2755
We've released @BananaMindBot .
Most things you do on HuggingFace, BananaMindBot can do. Fast
Mention @BananaMindBot on a model, dataset, Space discussion, paper, blog comment, or top-level post and it'll reply there.
It's powered by North Code Mini (Qwen3.8 27B, with GPT OSS 120B as fallback).
A few things it can do:
Search for models and datasets
Look up users and orgs and see what they've published
Read model cards, configs, dataset files, blog posts, and org profiles
Answer questions about what it finds
Write and run its own code in a locked-down sandbox when it needs to verify something
Check things like a model's real parameter count from the safetensors headers instead of just repeating the model card
Remember something for later if you explicitly ask it to
Forward a message to @Banaxi-Tech
Post a daily roundup of developments in the small-language-model space
It won't execute code you give it. It can read and review that code, but anything it runs is code it wrote itself.
It also can't access private data or credentials.
Mention it somewhere.
It's going to also find this post!
(Some parts inspired by CompactBot and @CompactAI Follow them please)
Most things you do on HuggingFace, BananaMindBot can do. Fast
Mention @BananaMindBot on a model, dataset, Space discussion, paper, blog comment, or top-level post and it'll reply there.
It's powered by North Code Mini (Qwen3.8 27B, with GPT OSS 120B as fallback).
A few things it can do:
Search for models and datasets
Look up users and orgs and see what they've published
Read model cards, configs, dataset files, blog posts, and org profiles
Answer questions about what it finds
Write and run its own code in a locked-down sandbox when it needs to verify something
Check things like a model's real parameter count from the safetensors headers instead of just repeating the model card
Remember something for later if you explicitly ask it to
Forward a message to @Banaxi-Tech
Post a daily roundup of developments in the small-language-model space
It won't execute code you give it. It can read and review that code, but anything it runs is code it wrote itself.
It also can't access private data or credentials.
Mention it somewhere.
It's going to also find this post!
(Some parts inspired by CompactBot and @CompactAI Follow them please)
- 37 replies
DedeProGames
updated a
collection 16 days ago