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: packed nvfp4 / mxfp8 layers follow a module dtype cast (#17131)
04cd919 Daniel Han committed 12 days ago
Studio RAG: clip re-embedded chunks and embeddings API inputs with batched token counts (#17032)
0efc5fc Daniel Han committed 12 days ago
Studio: each model's compile bundle holds only its own artifacts (#17110)
d4f72be Daniel Han committed 12 days ago
Unsloth Studio: refuse Cosmos3 on diffusers 0.39 and 0.40 instead of failing at load or generate (#16745)
6c486bd Leo Borcherding committed 12 days ago
Unsloth Studio / Desktop: queue image generations from the Images page and the API (#15005)
3c51fdc Leo Borcherding committed 12 days ago
Studio llama.cpp: remember the positive reasoning budget probe per binary (#17031)
d705406 Daniel Han committed 12 days ago
Studio tool output: strip ANSI in plain runs and find the tail without splitting the whole text (#16998)
297a50c Daniel Han committed 12 days ago
Studio chat: skip the transcript read when no message recorded a context fit (#16995)
29d4733 Daniel Han committed 12 days ago
Studio downloads: skip the GGUF requirement lookup on scoped progress polls (#17028)
2b1391b Daniel Han committed 12 days ago
Studio /v1/embeddings: count tokens in one batch and serialize float rows with tolist (#17030)
5e3b9ad Daniel Han committed 12 days ago
Studio: load the chat runtime on use instead of in the startup bundle (#16999)
609ebee Daniel Han committed 12 days ago
Studio diffusion runs: keep a passing full-load verdict per unchanged bundle for the run listing (#17033)
2a6cc73 Daniel Han committed 12 days ago
Deep Research: keep a clean attachment inventory clean when inserting the terminal fallback (#16997)
fa31fb1 Daniel Han committed 12 days ago
Studio VLM conversion: fetch repo image files on a pool, settle rows in order (#17029)
4e260ac Daniel Han committed 12 days ago
Studio cached models: keep parsed README front matter per unchanged file (#17034)
46029e8 Daniel Han committed 12 days ago
unsloth chat: parse the streamed Markdown when a frame is drawn, not on every chunk (#16994)
8360b2c Daniel Han committed 12 days ago
Studio SSM kernels: read the loaded torch build instead of a torch-import child per load (#16996)
5965706 Daniel Han committed 12 days ago
Studio LTX-2.5: one denoiser lease per CFG request, upload the next streamed encoder layer while this one runs (#17123)
e168963 Daniel Han committed 12 days ago
Studio: keep fp32 timestep and RoPE tables in fp32 on single-file loads (#17142)
6705f48 Daniel Han committed 12 days ago
Studio: no strip behind Back to Hub over a background image (#17152)
00c11c1 Michael Han committed 12 days ago
Studio Qwen-Image-2.1: reuse an unchanged condition image's VAE latents across renders (#17092)
e1e4cc6 Daniel Han committed 12 days ago
MoE DDP negative control: accept the idle rank's missing gradient, whichever rank fails first (#17151)
a402f1a Daniel Han committed 12 days ago
Studio: streamed torchao weights no longer crash after a compile (Cannot swap t1 ... weakref) (#17141)
c948925 Daniel Han committed 12 days ago
Studio: skip QKV fusion when projections are kept packed (#17132)
da0df6a Daniel Han committed 12 days ago
GRPO PrefixGrouper: engage only when the model's forward applies prefix_seg_info (#17058)
7ad3923 Daniel Han committed 12 days ago
Unsloth Studio: train video-capable VLMs on a dataset's video instead of dropping it (#16965)
9882ccb Leo Borcherding committed 12 days ago
Studio: keep the lora_adapter_metadata strength of a rewritten LoRA file (#17088)
a8c0efb Daniel Han committed 12 days ago
Studio desktop: steady floating browser window on drag, resize and float (#17027)
ca5e22e Michael Han committed 12 days ago
Studio MiniMax-H3: reuse the conditioner output when only the seed of a reference / keyframe render changes (#17094)
c6aabe5 Daniel Han committed 12 days ago
Studio video: spool encoded stream chunks instead of holding a second raw clip (#17066)
87b7acb Daniel Han committed 12 days ago