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
Tests: plant uv.exe on Windows in the MLX repair minimal-PATH lookup test (#15995)
081c1c5 Daniel Han committed 22 days ago
Studio: load downloaded models with fewer Hugging Face requests (#15844)
8c3add5 Nilay committed 22 days ago
Studio: share one edge index for Recipe Studio node connection status (#15942)
3bee6d8 Daniel Han committed 22 days ago
Studio: stop the KV cache type following the last model to the next one (#15819)
1f5d258 Nilay committed 22 days ago
Studio: keep the app responsive as an MLX or Transformers load finishes (#15842)
1833dcf Nilay committed 22 days ago
Studio: nvfp4 safetensors twins: converter 0-dim compare, refuse a twin the installed torchao cannot unflatten (#15977)
1724d98 Daniel Han committed 22 days ago
Studio: stop re-uploading the whole chat while a reply streams (#15818)
30555d4 Nilay committed 22 days ago
Studio: test that escaped currency keeps long replies streaming fast (#15820)
4447544 Nilay committed 22 days ago
Studio: keep the loaded model if a GGUF download fails or is cancelled (#15846)
a1fa6e6 Nilay committed 22 days ago
Tests: make unsloth_cli/tests/test_start.py pass on Windows (#15987)
41c39c4 Daniel Han committed 22 days ago
Studio: restart a crashed GGUF model instead of showing it loaded (#15847)
565b2c1 Nilay committed 22 days ago
Studio: let unsloth start vibe start with smaller context windows (#15863)
db7c2e4 Nilay committed 22 days ago
Studio: keep the user's git identity and gh login in unsloth start pi (#15865)
ba24b21 Nilay committed 22 days ago
Tests: keep tqdm's monitor thread out of the ShareGPT no-threads count (#15992)
81c219b Daniel Han committed 22 days ago
Studio: stop Auto context from reserving a model's whole window on a Mac (#15843)
f33a2b0 Nilay committed 22 days ago
Studio diffusion: a first load downloads the weights before the small-host decision reads them (#15974)
16cfa8a Daniel Han committed 22 days ago
Studio: match linked-folder paths against .gitignore runs, not rule by rule (#15949)
d0cc194 Daniel Han committed 22 days ago
Studio: answer dense RAG search from an in-memory numpy matrix kept in sync by SQLite triggers (#15961)
e5d3613 Daniel Han committed 22 days ago
Studio: encode each recall query once per compaction and let archive widening widen (#15960)
8cfa93f Daniel Han committed 22 days ago
Studio: say when remote access Start automatically opens a public URL (#15849)
782e786 Nilay committed 22 days ago
CI: split Repo tests (CPU, rest) and (CPU, gloo) before they reach their 2100 s cap (#15951)
4a87de1 Daniel Han committed 23 days ago
Studio video: MiniMax-H3 unfused Q / K / V share one ConvRot rotation and act quant (#15956)
f6e47fe Daniel Han committed 23 days ago
Studio video: MiniMax-H3 conditioner pin yields to a streamed denoiser on a small host (#15955)
2a7e3a5 Daniel Han committed 23 days ago
Tests: mac tab walk waits for the sidebar to mount before judging a pinned row (#15959)
37509ce Daniel Han committed 23 days ago
Studio: skip the audio gallery prune's sidecar pass under both caps and list prepared inputs once per sweep (#15874)
df42bc0 Daniel Han committed 23 days ago
Studio: encode a batch's gallery PNGs concurrently (#15909)
66ec3fc Daniel Han committed 23 days ago
Tests: read _gpu_init.py as UTF-8 in the torchao-too-new test (#15957)
49582cd Daniel Han committed 23 days ago
Studio diffusion: run both guidance branches per streamed block upload for Qwen-Image-Edit under a VRAM budget (#15930)
9c30724 Daniel Han committed 23 days ago