Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.
Python agentaichatgptdeepseekfine-tuninggemmaimage-generationllamallmllmsopenaipythonqwenreinforcement-learningself-hostedstable-diffusiontext-to-speechttsuiunsloth
Commits on main
Baseline the unsloth-zoo 2026.10.6 and transformers 5.20.0 scan findings after review (#13949)
ec411c3 Daniel Han committed 1 month ago
Studio: fix Desktop crash at launch on Wayland with NVIDIA and an iGPU (#13892)
061d525 Indrajit swain committed 1 month ago
NVFP4 tests: real-CUDA skip guards, and format the files #13840 left off the hook (#13937)
565630c Daniel Han committed 1 month ago
Fix main's red CI: UI font tokens in the token preview, real-CUDA skip guards (#13936)
6d7afe0 Daniel Han committed 1 month ago
NVFP4: decode weight tiles inside the GEMM for small-batch forward and dX (#13840)
fa15e69 Daniel Han committed 1 month ago
Studio: do not re-send the image generate POST after a dropped connection (#13790)
0737d7d Younes Beriane committed 1 month ago
Locate the MTP drafter-drop fallback by its shape, not one spelling (#13922)
bd06376 Daniel Han committed 1 month ago
Studio: preview tokens and the completion mask before training (#13850)
a2003f1 Daniel Han committed 1 month ago
Bump install.sh / install.ps1 pins to unsloth>=2026.10.6, unsloth-zoo>=2026.10.6 (#13918)
221d57b Daniel Han committed 1 month ago
Fix main's red CI after #13806, #13899 and recent formatter drift (#13909)
5e55110 Daniel Han committed 1 month ago
Fix main's red Backend CI, repo and picker contract tests after #13724, #13749 and #13828 (#13846)
d303bac Daniel Han committed 1 month ago
Studio: let Run settings load a Hugging Face model in 16-bit (#13841)
bebba39 Daniel Han committed 1 month ago
MoE LoRA: keep native per-expert Linears for an explicit target_modules list on transformers 4.x (#13851)
ddbbd9b Daniel Han committed 1 month ago
Studio: support NVIDIA Nemotron 3.5 Bolt and Super GGUFs (#13899)
4494fa5 Daniel Han committed 1 month ago
Frontend CI: keep grouped browser logs through a second cancel signal (#13894)
c4eb29a Daniel Han committed 1 month ago
Studio: train a speculative decoding drafter with fine-tunes and export it for llama.cpp and vLLM (#13806)
1be7795 Daniel Han committed 1 month ago
Studio: Windows torch 2.12 kernels from our prebuilt wheels (#13808)
dac778c Daniel Han committed 1 month ago
Fix FastLanguageModel fused CE loss under FSDP2 with an untied lm_head (#13827)
12288e5 Daniel Han committed 1 month ago
Studio Hub: show the Meta logo on every Unsloth Muse re-upload (#13896)
cd42465 Michael Han committed 1 month ago
Launch eager RMSNorm and LayerNorm kernels through triton_launch (#13889)
b67f20f Daniel Han committed 1 month ago
MoE grouped GEMM: tl.range for the dX expert loop, drop the dead dW FLATTEN (#13831)
d64b080 Daniel Han committed 1 month ago
Studio: serve /v1/messages and count_tokens from MLX and other in-process models (#13795)
9a9a4af Long Yixing committed 1 month ago
Stream the chunked cross entropy forward logits with evict_first (#13893)
945e2fc Daniel Han committed 1 month ago
FP8 rowwise GEMV: int64 row offsets past 2^31 elements, and skip it on Blackwell with Triton 3.3 (#13830)
06d9853 Daniel Han committed 1 month ago
NF4 GEMV: drop K masks on even K and gather the NF4 table from registers (#13829)
2ce16b5 Daniel Han committed 1 month ago