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: stop web results crowding attached files out of Deep Research (#17434)
ce489c1 Nilay committed 9 days ago
Studio: ask before Python runs a shell command passed by reference (#17456)
6fef7b1 Nilay committed 9 days ago
Studio: keep chat text out of the header with Glass or a background (#17429)
4bd4ea5 Nilay committed 9 days ago
Studio tests: give the llama-server owner tests a wait cap a loaded runner can meet (#17488)
9dc37ed Daniel Han committed 9 days ago
Flex attention decode steps for Gemma 4 and Qwen3.5 / Qwen3.6 under decode CUDA graphs (#17329)
db12e18 Daniel Han committed 9 days ago
Studio: MiniMax-H3-Image text to image and reference edits (#17453)
1f68ed5 Daniel Han committed 9 days ago
Tests: clear the cached tensor parallel layout before every test (#17472)
4b10969 Daniel Han committed 9 days ago
Studio NVFP4: 4over6 block scales at load and in the prequant builder (#17468)
952f49f Daniel Han committed 9 days ago
Studio: neutral NVFP4 gate field names and refreshed checkpoint hashes (#17469)
4e17a32 Daniel Han committed 9 days ago
Studio NVFP4: tune mm_fp4 once per M bucket and keep the tactics on disk (#17467)
116a955 Daniel Han committed 9 days ago
Studio: stop Reduce motion turning on a transition for every element (#17466)
e4ef631 Daniel Han committed 9 days ago
Studio sd.cpp: pin the prebuilt resynced with upstream master-1056 (#17459)
57a486b Daniel Han committed 9 days ago
Studio Terminal: give the heredoc quote scanner its own substitution helper (#17442)
322c4c1 Daniel Han committed 10 days ago
Unsloth Studio: show which models have saved run settings and reset them from the model menu (#12608)
22812d8 Leo Borcherding committed 10 days ago
Studio tests: read video.py as UTF-8 in the H3 fused SwiGLU test (#17437)
011129b Daniel Han committed 10 days ago
Studio: ask before running Python fed to Terminal through a heredoc (#17256)
1393084 Nilay committed 10 days ago
Update idempotency: let the offline test survive a version bump that is not on PyPI yet (#17428)
03c030d Daniel Han committed 10 days ago
Studio MiniMax-H3: feed the SwiGLU halves straight into the down projection's int8 activation quant (L4 -0.85% s/step, bit-identical) (#17422)
150d223 Daniel Han committed 10 days ago
Studio: keep more of a streamed FLUX.3 klein int8 transformer resident at small budgets (L4 6 GiB -4.2% per image, H2D 56.9 -> 43.9 GiB) (#17419)
77e48e3 Daniel Han committed 10 days ago
Scan baseline: re-review the seven unsloth-zoo sites 2026.11.4 moved (#17421)
a79e6d5 Daniel Han committed 10 days ago
Studio FLUX.3: keep float16 on fp16-only GPUs behind a residual rescale (T4 INT8 -24% per image, GGUF Q8_0 -73%) (#17385)
dd43f10 Daniel Han committed 10 days ago
Studio: mount the API monitor inside a host that stays mounted, so opening it in Glass stops restyling the app (#17418)
817a00c Daniel Han committed 10 days ago
Installer: clear damaged uv cache entries before retrying an install (#17398)
0dce385 Daniel Han committed 10 days ago
Installers: require unsloth-zoo>=2026.11.4 to match the pyproject floor (#17414)
988ae96 Daniel Han committed 10 days ago
Studio: leave room for thinking in a Custom connection's chat title request (#17332) (#17363)
b23f373 Indrajit swain committed 10 days ago
Studio: set scrollbar-color in rgb so Reduce motion stops restarting a transition on every element (#17413)
b44acd8 Daniel Han committed 10 days ago