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
studio: fetch the XPU triton wheel with --dest, not pip's -d (#10073)
32dcc75 Wayne Sletcher committed 2 months ago
studio: use generation usage for conversation search budgets (#9438)
6e21a68 Maheswar Kumar committed 2 months ago
Studio: keep tools available when a picture is attached (#10092)
97e12a2 Nilay committed 2 months ago
Fix stale Qwen sampling defaults after Studio upgrades (#9643)
96886fa Lee Jackson committed 2 months ago
Revert "Studio: auto-generate an admin password for headless public launches (#7392)"
9ca3a44 oobabooga committed 2 months ago
Studio: keep download restart toasts job-scoped (#10136)
7466107 Wasim Yousef Said committed 2 months ago
Studio: line up On Device rows, and list partial downloads (#10128)
daf0c80 Michael Han committed 2 months ago
Studio: restore Deep Research plan approval (#10134)
71b7b30 Wasim Yousef Said committed 2 months ago
Studio: keep one MCP connection per chat instead of reconnecting every call (#10095)
0c0e45e Nilay committed 2 months ago
Studio: auto-generate an admin password for headless public launches (#7392)
6c07765 Daniel Han committed 2 months ago
studio: fall back to the MTP/ folder when a repo publishes no root drafter (#10124)
9d4ee61 Daniel Han committed 2 months ago
Stop `unsloth start claude` from opening with a model it cannot use (#10089)
c59cd2d Nilay committed 2 months ago
Bump the cargo-tauri group across 1 directory with 6 updates (#9798)
47421f3 dependabot[bot] committed 2 months ago
feat(studio): give MLX models a real context length, from the model config to the KV cache (#8125)
a09bb58 Long Yixing committed 2 months ago
Make the notebooks smoke job reach the check it exists for (#10125)
31af9ab Daniel Han committed 2 months ago
Cache the same Playwright engines under the same key (#10122)
ddd22ae Daniel Han committed 2 months ago
Prune an open PR's caches once nothing has read them (#10123)
30261dc Daniel Han committed 2 months ago
Studio: use plain arrow icons for up, down, left and right (#10129)
18cbcdb Michael Han committed 2 months ago
fix(AMD): route Fedora/Bazzite to per-arch ROCm wheels when no version is readable (#9152)
0833157 Leo Borcherding committed 2 months ago
fix(amd): don't enable torch.compile when Triton's C toolchain can't compile it (Windows) (#7704)
607a8af Leo Borcherding committed 2 months ago
Kaggle T4: split the load phase into fetch and weight load (#9692)
e18a069 Daniel Han committed 2 months ago