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: size more audio.cpp music models from measured peaks and show the load estimate in the picker (#16726)
dcdff7c oobabooga committed 15 days ago
Studio: quantize Qwen-Image-2.1-Turbo NVFP4 with the Qwen-Image-2.1 layer policy (#16757)
e2daf16 Daniel Han committed 15 days ago
perf(studio): stream MLX VLM SentencePiece tokens as they are sampled (#16684)
a34d766 Long Yixing committed 15 days ago
Studio: fix two first-use regressions from deferred startup hardware detection (#16629)
23907ad Daniel Han committed 15 days ago
Studio: load Anima, Chroma, MOSS-Image and int4 Krea 2 ComfyUI single files (#16710)
b7a0a85 oobabooga committed 15 days ago
Studio: read adapter_config.json once per inference worker load (#16604)
c006ec3 Daniel Han committed 15 days ago
Studio: keep cancelled fine-tuning jobs in paginated results (#16633)
8e0a102 Max-Reisinger committed 15 days ago
perf(studio): queue MLX prompt-cache snapshots without waiting on the GPU (#16683)
305afc9 Long Yixing committed 15 days ago
Studio: judge an HSA_OVERRIDE_GFX_VERSION GPU by its override in the cold engine status (#16670)
c9f786f Daniel Han committed 15 days ago
Studio: browser dictation in Chrome no longer repeats the last word (#16685)
3561a41 Nilay committed 15 days ago
Studio RAG: reuse one rag.db connection per thread within a linked-folder sync pass (#16607)
f240d74 Daniel Han committed 15 days ago
Studio remote-code scan: read each Hub config once per scan request (#16606)
af173fc Daniel Han committed 15 days ago
Studio: read each thread's stored messages once per chat re-import (#16615)
79bf94e Daniel Han committed 15 days ago
Studio: list a GGUF repo once for the mmproj, MTP, DSpark and DFlash lookups of one load (#16565)
7ddad80 Daniel Han committed 15 days ago
Studio: skip the NVML memory probe child when CUDA_VISIBLE_DEVICES is blank (#16597)
f609b45 Daniel Han committed 15 days ago
Studio MLX images: faster GGUF to MLX bit packing, byte-identical output (#16746)
2a2cc59 Daniel Han committed 15 days ago
Studio: the in-app browser can float over the chat as a small, resizable window (#16775)
4db1f39 Michael Han committed 15 days ago
Studio: fingerprint the same manifest scopes the downgrade migration scan reads (#16556)
94f23e0 Daniel Han committed 15 days ago
Desktop: keep the bundled AppImage tray loader off the host's Ayatana (#16613)
0d9d7a5 Debojyoti Banerjee committed 15 days ago
Studio: search for llama-server once per Decision API settings answer (#16617)
a6c7e49 Daniel Han committed 15 days ago
GRPO: keep packed forwards off for models that mix tokens across packed rows (#16349)
4624f4a Shubhankar Tripathy committed 15 days ago
open_login test: bring its own Playwright error types so a leaked stub cannot break it (#16784)
2395605 Daniel Han committed 15 days ago
Studio: floating monitors stay under the pointer and land without a flash (#16781)
04f4f0a Michael Han committed 15 days ago
Add opt-in BF16 AdamW with stochastic parameter updates (#16662)
f8269bd Daniel Han committed 15 days ago
Studio: rename a managed account without recreating it (#16668)
5b2ef87 Daniel Han committed 15 days ago
Studio: let an approved Python call reach a host outside the sandbox allowlist (#16674)
e60ca31 Daniel Han committed 15 days ago