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
Long-chat gate coverage: measure forced motion now that Reduce motion is on by default (#17395)
baf518b Daniel Han committed 10 days ago
Studio: open update-notes links in the Desktop app, treat .term and .tcsh downloads as code (#17394)
358e8d1 Daniel Han committed 10 days ago
Studio: edit_file and Rewind write through an open chat-folder descriptor on POSIX (#17389)
43625c9 Daniel Han committed 10 days ago
Studio: do not start a queued fine-tuning job for a disabled or deleted account (#17392)
ac8c5cd Daniel Han committed 10 days ago
Studio: cap request bodies on every route and decode zstd bodies one at a time (#17388)
d2a69b7 Daniel Han committed 10 days ago
AsyncGRPO: require a per-run secret on the local vLLM server; validate BF16 AdamW state on load (#17390)
41bee13 Daniel Han committed 10 days ago
Decision models on MLX: refuse a config.json model_file unless trusted (#17391)
3144ae6 Daniel Han committed 10 days ago
Studio: end plain-text training rows with EOS when Format is Auto (#17255)
58973e8 Nilay committed 10 days ago
Studio: hold the login rate limit on the Colab public link (#17259)
a64065f Nilay committed 10 days ago
Studio: stop audio models learning to answer None for blank transcripts (#17257)
fa115cd Nilay committed 10 days ago
Decode CUDA graphs for Qwen3.5 / Qwen3.6 hybrid linear-attention caches (#17324)
7c32794 Daniel Han committed 10 days ago
Studio: save LoRA and merged exports of a run to separate folders (#17260)
7caea9b Nilay committed 10 days ago
Studio: keep hidden spreadsheet rows in chat attachments (#17314)
457795e Nilay committed 10 days ago
Slot-clamp lint: accept the allow marker on any line of a wrapped assignment (#17384)
db0367a Daniel Han committed 10 days ago
Studio MiniMax-H3: run one denoiser forward per requested step (20 steps ran 19) (#17362)
8699f88 Daniel Han committed 10 days ago
Studio: keep a user-set Hermes api_key in unsloth start hermes (#17320)
b71ee2a Nilay committed 10 days ago
Skip the duplicate tokenizer load on transformers 5 when it can only copy the fast tokenizer (#17318)
4fd58f0 Daniel Han committed 10 days ago
Studio: keep the Ask before downloading programs setting across browsers (#17240)
3d317d4 Nilay committed 10 days ago
Studio: read once per tab return instead of on both visibility and focus (#17300)
4189fa2 Daniel Han committed 10 days ago
Studio installers: ask for the latest llama.cpp / whisper.cpp release tag once per run (#17289)
9bd6cd6 Daniel Han committed 10 days ago
Studio Windows GPU inventory: run the PowerShell adapter scan only when AMD or Intel records need it (#17272)
59244b6 Daniel Han committed 10 days ago
Studio: keep searching the web past a lone Wikipedia result (#17242)
100352c Nilay committed 10 days ago
Studio: keep speaker notes when a PowerPoint file is attached in chat (#17258)
5180805 Nilay committed 10 days ago
Studio: resolve the nested Radix primitive / slot / context copies to one version (#17301)
a0f89da Daniel Han committed 10 days ago
Diffusion guard test: match the calibrated placement and ControlNet kwargs structurally (#17376)
9947de2 Daniel Han committed 10 days ago
Studio Anima LoRA: rule out kohya splits that name no registered module before probing (#17247)
0036e84 Daniel Han committed 10 days ago
Studio: cache the native-template markup profile per tokenizer in renderable_tool_catalog (#17288)
ddba6f7 Daniel Han committed 10 days ago