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: load the retired FLUX.3 full copies through their base (#14173)
83dc7f8 Daniel Han committed 1 month ago
Studio: MCP agents upload datasets, load what they trained, and download models ahead of time (#14107)
21f02a8 Etherl committed 1 month ago
Studio: MCP agents choose how and where a model loads, preview the load, and queue videos (#14090)
21e201b Etherl committed 1 month ago
Studio: show the running step on the Hub card's download bar (#14171)
93bd68f Michael Han committed 1 month ago
Studio: choose Off, When they fit or Always in Model memory, and keep the chat context when an image or video model takes the GPU (#14108)
22531d3 Etherl committed 1 month ago
Studio: MCP agents follow long work in one call and see every model a load unloads (#14089)
e2a214b Etherl committed 1 month ago
Studio: MCP errors, cancels and chat controls agents can act on (#14081)
43ceef8 Etherl committed 1 month ago
Studio: image and video models stay loaded beside chat when they fit (#14106)
9bac178 Etherl committed 1 month ago
Studio: add an Always choice to Keep multiple models loaded, track every model on the GPU, and report every model a load unloads (#14088)
9edc731 Etherl committed 1 month ago
Studio: MCP returns key-free download links and finds media, voices and video frames again (#14079)
904a615 Etherl committed 1 month ago
Studio: the Images pill makes images and videos in chat with the loaded local models (#14110)
56d1204 Etherl committed 1 month ago
Studio: MCP audio reports what it applied, explains audio errors, checks takes and cancels by id (#14080)
8252333 Etherl committed 1 month ago
Studio: add a setting to stop chats and Deep Research when the tab closes (#14062)
8e69f20 alkinun committed 1 month ago
GRPO: latch autocast under vLLM only, zero-variance exclusion on TRL before 1.6 (#14123)
8e3fd96 Daniel Han committed 1 month ago
Studio: add Edit image to the Images gallery item menu (#14162)
b52d34a Michael Han committed 1 month ago
Studio: drop an Images strip tile onto an image picker to use it (#14166)
4531eb5 Michael Han committed 1 month ago
Studio: image and video generation follow each model's defaults and prompting tips, and /v1/videos takes Unsloth Studio controls (#14078)
4a1037e Etherl committed 1 month ago
Studio: clearer audio.cpp errors, cleaner Speak/Clone, reproducible takes (#14075)
d18ecfc Etherl committed 1 month ago
Studio: one-command MCP setup, recipe prompts, a model catalog and tool annotations (#14077)
4bed2fa Etherl committed 1 month ago
Studio: keep the download button off the chat composer on phones (#14163)
f2ed007 Michael Han committed 1 month ago
Tests: stop a GPU-selection route test leaking a CHAT owner into the upsampler tests (#14169)
2f8b62c Daniel Han committed 1 month ago
Log budget: measure the FLUX.3 prompt upsampler status poll (#14167)
d67eba9 Daniel Han committed 1 month ago
AMD install follow-ups: installable inferred target, gfx1153 lspci peer, override health hint (#14117)
6069a27 Daniel Han committed 1 month ago
RMSNorm: keep the final norm's own dX buffer when hidden states are returned to the caller (#14139)
7462b48 Daniel Han committed 1 month ago
Studio: fetch every video component from unsloth repos (#14003)
3111a2c Daniel Han committed 1 month ago
Studio: image families default to a hosted quantized text encoder (#14009)
8ef162f Daniel Han committed 1 month ago
Studio: fetch every image component from unsloth repos (#14008)
a6da154 Daniel Han committed 1 month ago