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 engines: verify TLS against Studio's certifi bundle (#15509)
1cb33f7 Daniel Han committed 25 days ago
fix(studio): price DeepSeek-V4.1 oQ quants as unsloth-zoo loads them (#15508)
e798a57 Long Yixing committed 25 days ago
Studio: track MPS denoise progress with torch.mps events (#15323)
f2bc9d5 Long Yixing committed 25 days ago
Keep each unsloth eval run's results instead of overwriting the last (#15451)
d69407d Nilay committed 25 days ago
Studio: update managed Node to 24.22.1, apply new pins on every update path, and keep the pin current (#15423)
4c37b87 oobabooga committed 25 days ago
Studio: start overlapping document chunks on a whole word (#15442)
abaa1b5 Nilay committed 25 days ago
fix(studio): time a speculative vision batch row's prefill from before add() (#15517)
b6fc821 Long Yixing committed 25 days ago
Studio: respell user ngram-mod extra args for older llama.cpp builds (#14880)
b04b740 Daniel Han committed 25 days ago
MXFP4 dequantize branch test: run the locals the branch reads ahead of it (#15558)
3c1a0b8 Daniel Han committed 25 days ago
Make the one scored completion GRPO test follow the TRL version (#14775)
cd7850c Daniel Han committed 25 days ago
Studio: keep a temporary chat's browser agent pages out of history (#14956)
3bfc786 Daniel Han committed 25 days ago
Studio diffusion: store chunked int8 W8A8 results straight into the output rows (#15533)
5541b47 Daniel Han committed 25 days ago
Studio: thinking support for models served by the SGLang engine (#15433)
1c6b24e Nilay committed 25 days ago
Studio: queue OpenAI fine-tuning jobs while the GPU is busy (#14645)
4d5e78c Daniel Han committed 25 days ago
Studio native cache write: read each weight from the live module at write time so offload can reclaim VRAM (#15461)
a2e6390 Daniel Han committed 25 days ago
Studio: refuse a reused temp folder that other users can still read (#14962)
91d9280 Daniel Han committed 25 days ago
cuDNN SDPA backward: give grad_output the output's layout on torch 2.6 - 2.13 (#15445)
2df2860 Daniel Han committed 25 days ago
Studio diffusion: keep reference headroom on multi-GPU FLUX.2 klein splits so edits are not refused (#14858)
c9af351 Daniel Han committed 25 days ago
Studio diffusion: fused RMSNorm on the fp16 eager Z-Image path (#15543)
d3e62a9 Daniel Han committed 25 days ago
Refuse Trainer(model_init=...) under UNSLOTH_TP_SIZE instead of silently skipping tensor parallel (#15424)
f5d14f5 Daniel Han committed 25 days ago
Studio: keep browser tabs clickable on macOS with Safari 18.4 and earlier (#14368)
2e973ce Daniel Han committed 25 days ago
Block compile: mark Unsloth's own compiled region apart from a user's enclosing compile (#15526)
73560d3 Daniel Han committed 25 days ago
QK norm + RoPE backward: pin the dX contraction so compiled fp16 matches eager on ROCm (#15506)
ddd9838 Daniel Han committed 25 days ago
Studio video: MiniMax-H3 cache-only loads name the missing file and how to fix it (#15547)
f55886c Daniel Han committed 25 days ago
Studio training: describe public mirror access separately from model licensing (#15475)
0c9ac80 Daniel Han committed 25 days ago
Studio video: drain palettegen per frame in the two-pass GIF export (#14958)
9e88734 Daniel Han committed 25 days ago
Studio: size MIG CUDA instances from their own slice, never a physical card row (#15465)
89a08ba Daniel Han committed 25 days ago
Studio: revealing a chat's sandbox folder is owner-only (#15512)
ca29b72 Daniel Han committed 25 days ago