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
Tests: stop pytest 9.2 monkeypatch from deleting torch config entries, and name any test that still does (#15626)
fc03e53 Daniel Han committed 24 days ago
Studio: ask before Clear all deletes the desktop transcript history (#15570)
4e673d3 Nilay committed 24 days ago
Persist chat pins across clients of the same Unsloth server (#15293)
71dd6d2 Lee Jackson committed 24 days ago
Studio: leave room for the whole reply cap in the estimated managed-engine fit (#15610)
a665aa0 Daniel Han committed 24 days ago
gpt-oss: refuse fused 4-bit with dtype float32 on GPUs without bfloat16 (#15614)
e55893c Daniel Han committed 24 days ago
Studio video: price deferred pins at their bytes so a returning Wan load pins its text encoder (#15612)
702378d Daniel Han committed 24 days ago
Studio: chat with a fine-tuned model by the name its job returns (#15449)
c4f9de9 Nilay committed 24 days ago
fix(studio): make room for the MLX drafter in the auto offload plan (#15342)
6322f7b Long Yixing committed 24 days ago
cuDNN SDPA DTensor test: gate on a real CUDA device (#15582)
b3c7886 Daniel Han committed 24 days ago
Studio diffusion: pin the up-front slabs in registered host memory (faster load, same bytes) (#15562)
86f8bd2 Daniel Han committed 24 days ago
Keep Codex MCP servers when running unsloth start codex (#15452)
d0349f0 Nilay committed 24 days ago
Studio: resume cut-off replies and structured output on vLLM and SGLang (#15435)
15569be Nilay committed 24 days ago
Studio: keep linked folders syncing when files become ignored (#15441)
4280227 Nilay committed 24 days ago
Studio: keep the MLX chat context when an image or video model takes the GPU (#15358)
20f3ac0 Long Yixing committed 25 days ago
GRPO kept-set refusals: drop the automatic-default wording (#15531)
c8cef55 Daniel Han committed 25 days ago
Studio: train compatible queued fine-tuning jobs together with multi-LoRA (#15344)
3e3c2fd Daniel Han committed 25 days ago
Studio video: pin MiniMax-H3's reduction-config filter (same seed, same clip across processes) (#15564)
2459c93 Daniel Han committed 25 days ago
Studio video: map MiniMax-H3's hosted int8 checkpoint instead of reading it whole (faster load, same weights) (#15563)
739bdfa Daniel Han committed 25 days ago
Studio: keep documents searchable after upgrading a Chat only install (#15443)
8e1fb9a Nilay committed 25 days ago
Studio mermaid guard: a quoted @{ cannot hide a later remote image (#14907)
ca59a0d Daniel Han committed 25 days ago
Studio: enable packing and assistant-only training defaults where supported (#15478)
e047171 Daniel Han committed 25 days ago
feat(studio): store ComfyUI fp8 layers as fp8 on Apple MPS when the widened model would not fit (#15507)
d742d07 Long Yixing committed 25 days ago
Studio: keep sidebar sections and favorites made in another window (#15440)
1814733 Nilay committed 25 days ago
Studio: keep chat-installed packages out of plugin discovery on approved host calls (#14948)
6ef7574 Daniel Han committed 25 days ago