Scaling Test-Time Compute for Agentic Coding
Paper β’ 2604.16529 β’ Published β’ 10
AI & ML interests
Unofficial org for community upload of Mistral's Open Source models.
Recent Activity
View all activity
nielsrΒ
submitted a
paper to Daily Papers 5 days ago
danielhanchenΒ
posted an update 6 days ago
Post
5170
Qwen3.6-27B is out now! Run it locally on 18GB RAM. π
Qwen3.6-27B surpasses Qwen3.5-397B-A17B on all major coding benchmarks.
GGUFs to run: unsloth/Qwen3.6-27B-GGUF
Guide + MLX: https://unsloth.ai/docs/models/qwen3.6
Qwen3.6-27B surpasses Qwen3.5-397B-A17B on all major coding benchmarks.
GGUFs to run: unsloth/Qwen3.6-27B-GGUF
Guide + MLX: https://unsloth.ai/docs/models/qwen3.6
danielhanchenΒ
posted an update 12 days ago
Post
2777
Qwen3.6-35B-A3B can now be run locally! π
The model is the strongest mid-sized LLM on nearly all benchmarks.
Run on 23GB RAM via Unsloth Dynamic GGUFs.
GGUFs to run: unsloth/Qwen3.6-35B-A3B-GGUF
Guide: https://unsloth.ai/docs/models/qwen3.6
The model is the strongest mid-sized LLM on nearly all benchmarks.
Run on 23GB RAM via Unsloth Dynamic GGUFs.
GGUFs to run: unsloth/Qwen3.6-35B-A3B-GGUF
Guide: https://unsloth.ai/docs/models/qwen3.6
- 13 replies
nielsrΒ
submitted a
paper to Daily Papers 12 days ago
nielsrΒ
submitted a
paper to Daily Papers 19 days ago
danielhanchenΒ
posted an update 21 days ago
Post
5410
You can now fine-tune Gemma 4 for free with our notebooks! π₯
You just need 8GB VRAM to train Gemma 4 locally!
Unsloth trains Gemma4 1.5x faster with 50% less VRAM.
GitHub: https://github.com/unslothai/unsloth
Guide + Notebooks: https://unsloth.ai/docs/models/gemma-4/train
You just need 8GB VRAM to train Gemma 4 locally!
Unsloth trains Gemma4 1.5x faster with 50% less VRAM.
GitHub: https://github.com/unslothai/unsloth
Guide + Notebooks: https://unsloth.ai/docs/models/gemma-4/train
- 5 replies
nielsrΒ
submitted a
paper to Daily Papers 25 days ago
danielhanchenΒ
posted an update 26 days ago
Post
3773
Google releases Gemma 4. β¨
Gemma 4 introduces 4 models: E2B, E4B, 26B-A4B, 31B.
The multimodal reasoning models are under Apache 2.0.
Run E2B and E4B on ~6GB RAM, and on phones. Run 26B-A4B and 31B on ~18GB.
GGUFs: https://huggingface.co/collections/unsloth/gemma-4
Guide: https://unsloth.ai/docs/models/gemma-4
Gemma 4 introduces 4 models: E2B, E4B, 26B-A4B, 31B.
The multimodal reasoning models are under Apache 2.0.
Run E2B and E4B on ~6GB RAM, and on phones. Run 26B-A4B and 31B on ~18GB.
GGUFs: https://huggingface.co/collections/unsloth/gemma-4
Guide: https://unsloth.ai/docs/models/gemma-4
danielhanchenΒ
posted an update 28 days ago
Post
2740
A new way to use Unsloth.
Coming soon...
Coming soon...
MaziyarPanahiΒ
posted an update 28 days ago
Post
1883
Training mRNA Language Models Across 25 Species for $165
We built an end-to-end protein AI pipeline covering structure prediction, sequence design, and codon optimization. After comparing multiple transformer architectures for codon-level language modeling, CodonRoBERTa-large-v2 emerged as the clear winner with a perplexity of 4.10 and a Spearman CAI correlation of 0.40, significantly outperforming ModernBERT. We then scaled to 25 species, trained 4 production models in 55 GPU-hours, and built a species-conditioned system that no other open-source project offers. Complete results, architectural decisions, and runnable code below.
https://huggingface.co/blog/OpenMed/training-mrna-models-25-species
We built an end-to-end protein AI pipeline covering structure prediction, sequence design, and codon optimization. After comparing multiple transformer architectures for codon-level language modeling, CodonRoBERTa-large-v2 emerged as the clear winner with a perplexity of 4.10 and a Spearman CAI correlation of 0.40, significantly outperforming ModernBERT. We then scaled to 25 species, trained 4 production models in 55 GPU-hours, and built a species-conditioned system that no other open-source project offers. Complete results, architectural decisions, and runnable code below.
https://huggingface.co/blog/OpenMed/training-mrna-models-25-species
danielhanchenΒ
posted an update about 1 month ago
Post
922
You donβt need to set LLM parameters anymore! π
llama.cpp uses only the context length + compute your local setup needs. Unsloth also auto-applies the correct model settings
Try in Unsloth Studio - now with precompiled llama.cpp binaries.
GitHub: https://github.com/unslothai/unsloth
llama.cpp uses only the context length + compute your local setup needs. Unsloth also auto-applies the correct model settings
Try in Unsloth Studio - now with precompiled llama.cpp binaries.
GitHub: https://github.com/unslothai/unsloth
- 2 replies
MaziyarPanahiΒ
posted an update about 1 month ago
Post
2245
We annotated 119K medical images with two frontier VLMs (Qwen 3.5, Kimi K2.5), cross-validated at 93% agreement, and produced 110K training records, all for under $500. Fine-tuning 3 small models (2-3B params) improved all benchmarks: best model reaches +15.0% average exact match.
Everything is open-sourced: datasets, adapters, and code.
https://huggingface.co/blog/OpenMed/synthvision
Everything is open-sourced: datasets, adapters, and code.
https://huggingface.co/blog/OpenMed/synthvision
- 2 replies
nielsrΒ
submitted 3
papers to Daily Papers about 1 month ago
Do VLMs Need Vision Transformers? Evaluating State Space Models as Vision Encoders
Paper β’ 2603.19209 β’ Published β’ 5
V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learning
Paper β’ 2603.14482 β’ Published β’ 32
Omnilingual MT: Machine Translation for 1,600 Languages
Paper β’ 2603.16309 β’ Published β’ 21
danielhanchenΒ
posted an update about 1 month ago
Post
3401
Introducing Unsloth Studio β¨
A new open-source web UI to train and run LLMs.
β’ Run models locally on Mac, Windows, Linux
β’ Train 500+ models 2x faster with 70% less VRAM
β’ Supports GGUF, vision, audio, embedding models
β’ Auto-create datasets from PDF, CSV, DOCX
β’ Self-healing tool calling and code execution
β’ Compare models side by side + export to GGUF
GitHub: https://github.com/unslothai/unsloth
Blog and Guide: https://unsloth.ai/docs/new/studio
Available now on Hugging Face, NVIDIA, Docker and Colab.
A new open-source web UI to train and run LLMs.
β’ Run models locally on Mac, Windows, Linux
β’ Train 500+ models 2x faster with 70% less VRAM
β’ Supports GGUF, vision, audio, embedding models
β’ Auto-create datasets from PDF, CSV, DOCX
β’ Self-healing tool calling and code execution
β’ Compare models side by side + export to GGUF
GitHub: https://github.com/unslothai/unsloth
Blog and Guide: https://unsloth.ai/docs/new/studio
Available now on Hugging Face, NVIDIA, Docker and Colab.
danielhanchenΒ
posted an update about 2 months ago
Post
3927
We collaborated with NVIDIA to teach you about Reinforcement Learning and RL environments. π Learn:
β’ Why RL environments matter + how to build them
β’ When RL is better than SFT
β’ GRPO and RL best practices
β’ How verifiable rewards and RLVR work
Blog: https://unsloth.ai/blog/rl-environments
β’ Why RL environments matter + how to build them
β’ When RL is better than SFT
β’ GRPO and RL best practices
β’ How verifiable rewards and RLVR work
Blog: https://unsloth.ai/blog/rl-environments
- 4 replies
nielsrΒ
authored a
paper about 2 months ago
MaziyarPanahiΒ
posted an update about 2 months ago
Post
4848
DNA, mRNA, proteins, AI. I spent the last year going deep into computational biology as an ML engineer. This is Part I of what I found. π§¬
In 2024, AlphaFold won the Nobel Prize in Chemistry.
By 2026, the open-source community had built alternatives that outperform it.
That's the story I find most interesting about protein AI right now. Not just the science (which is incredible), but the speed at which open-source caught up. Multiple teams, independently, reproduced and then exceeded AlphaFold 3's accuracy with permissive licenses. The field went from prediction to generation: we're not just modeling known proteins anymore, we're designing new ones.
I spent months mapping this landscape for ML engineers. What the architectures actually are (spoiler: transformers and diffusion models), which tools to use for what, and which ones you can actually ship commercially.
New post on the Hugging Face blog: https://huggingface.co/blog/MaziyarPanahi/protein-ai-landscape
Hope you all enjoy! π€
In 2024, AlphaFold won the Nobel Prize in Chemistry.
By 2026, the open-source community had built alternatives that outperform it.
That's the story I find most interesting about protein AI right now. Not just the science (which is incredible), but the speed at which open-source caught up. Multiple teams, independently, reproduced and then exceeded AlphaFold 3's accuracy with permissive licenses. The field went from prediction to generation: we're not just modeling known proteins anymore, we're designing new ones.
I spent months mapping this landscape for ML engineers. What the architectures actually are (spoiler: transformers and diffusion models), which tools to use for what, and which ones you can actually ship commercially.
New post on the Hugging Face blog: https://huggingface.co/blog/MaziyarPanahi/protein-ai-landscape
Hope you all enjoy! π€
- 2 replies
merveΒ
updated a
Space about 2 months ago