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: whisper.cpp updates ride along with llama.cpp instead of opening the update card (#13862)
b24c27d Daniel Han committed 1 month ago
unsloth start openclaw: stop writing the legacy exec-approvals.json that OpenClaw 2026.8.1+ refuses to run with (#14343)
acc10af Daniel Han committed 1 month ago
Studio: send the TTS language to Supertonic and add a Speak language picker (#14511)
32491ae Daniel Han committed 1 month ago
Studio: opt-in REAP expert pruning on the Export page (#14130)
61fb544 Daniel Han committed 1 month ago
Tests: follow #14473's Unsloth wording in the sandbox, xformers and autoopt parity checks (#14522)
db0753a Daniel Han committed 1 month ago
Windows installer: restore the split-layout comment in setup.ps1 (Defender ML false positive) (#14520)
46c3bbb Daniel Han committed 1 month ago
FSDP2 gloo tests: stub every helper the trainer wrapper calls, not a fixed list (#14518)
ab26545 Daniel Han committed 1 month ago
CLI: --parallel help describes the shared KV cache (#14340)
4293241 Daniel Han committed 1 month ago
Copy uv constraint files to a space-free path on Windows volumes without 8.3 names (#14507)
deaf619 Daniel Han committed 1 month ago
Studio: route krea2 GGUFs to Images instead of opening them as chat models (#14477)
5ff3fa4 Indrajit swain committed 1 month ago
Studio diffusion: keep a model-offloaded denoiser's step eager instead of re-recording its CUDA graph every render (#14497)
4d7ebf5 Daniel Han committed 1 month ago
GRPO: opt-in rollout logprobs as the old policy (UNSLOTH_GRPO_ROLLOUT_LOGPROBS_AS_OLD) (#14489)
9f303b3 Daniel Han committed 1 month ago
Cross entropy: 16384 / 8 warp chunked forward on T4 and A100 (#14498)
0a20d7b Daniel Han committed 1 month ago
Dequantize FP8 weights in the activation dtype for fp16 LoRA (#14495)
b2b9852 Daniel Han committed 1 month ago
Keep has_internet from setting the process-wide socket timeout (#14378)
758d008 Vineeth Sai Varikuntla committed 1 month ago
Block compile: test that checkpoint early-stop does not demote the layer (#14065)
d477ae8 Indrajit swain committed 1 month ago
Tests: real-accelerator skip guards and utf-8 read in the gpt-oss and gc-skip tests (#14504)
c08dfbb Daniel Han committed 1 month ago
Studio: media decodes open no protocols, Data Recipe names stay inside their folder (#14463)
5a5aca2 Daniel Han committed 1 month ago
Studio: make CPU RAG embedding fast on Windows ROCm torch and honour Windows CPU affinity (#14354)
b2ccaa0 Daniel Han committed 1 month ago
Studio: login limiter and keyless access behind a reverse proxy (#14459)
71099f2 Daniel Han committed 1 month ago
Studio: view, edit and turn off chat memory in Settings > Chat (#14357)
dd5d02e Daniel Han committed 1 month ago
Unsloth Studio / Desktop: keep floating panels where they were left across resizes, minimize and restarts (#14355)
8a07b9c Daniel Han committed 1 month ago
Studio: MCP widget approval cards show the arguments being sent (#14458)
dff0fb9 Daniel Han committed 1 month ago
Studio: per-user temp folders, tighter tool sandbox grants, no UNC paths for managed accounts (#14461)
e3f5e60 Daniel Han committed 1 month ago
Studio: same diffusion pixels on every B200 server for FLUX.2 Klein and Z-Image (#14483)
5fdcfe5 Daniel Han committed 1 month ago
Studio: make Back in the browser panel work like a browser's (#14283)
a25776e Nilay committed 1 month ago
Studio diffusion: allow the multi-GPU layer split on discrete ROCm cards (#14479)
04d2bee Daniel Han committed 1 month ago