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: KV cache dtype, GPU memory limit and seed for vLLM and SGLang loads (#13596)
742a228 Daniel Han committed 1 month ago
DDP LoRA: default find_unused_parameters=False so gradient checkpointing does not crash (#13267)
0f5940b Daniel Han committed 1 month ago
Studio: pass Codex MCP and sub-agent tools to the model on /v1/responses (#13483)
0350034 Nilay committed 1 month ago
Studio: run Laya on the GPU by default and free it on unload (#13479)
c762842 Nilay committed 1 month ago
Studio: keep the local Claude agent from stopping on a small context window (#13477)
9cf63a2 Nilay committed 1 month ago
Studio: train and chat with every text-diffusion LM family (#13342)
2e718dc Daniel Han committed 1 month ago
Freeze the setup-time heap during Trainer.train(); add resume parity and memory-flat GPU tests (#13332)
abfe314 Daniel Han committed 1 month ago
FP8 on ROCm: use the dequant path on archs without e4m3fn Triton support (#13331)
eaad1bf Daniel Han committed 1 month ago
FP8 LoRA: dequantize down_proj in the backward like gate and up (#13270)
9e74eb1 Daniel Han committed 1 month ago
Re-measure the Studio startup budget after the audio selector and thinking merges (#13640)
cb6cbe8 Daniel Han committed 1 month ago
Studio: stop dropping the page text after a menu with no closing tag (#13489)
530c0da Nilay committed 1 month ago
Studio: vLLM thinking effort levels, preserve thinking and auto-compact (#13571)
8af1232 Daniel Han committed 1 month ago
Studio: hold the forced aligner against a folder delete; match compound audio.cpp size scopes in order (#13633)
89080f0 Daniel Han committed 1 month ago
tests: click the inline Delete button #13615 put back on the action bar (#13634)
03022a4 Daniel Han committed 1 month ago
Studio: keep ticked Word form check boxes in Chat with files (#13485)
1e87a42 Nilay committed 1 month ago
Studio: offer the precision folders of an MLX repo like GGUF quants (#13565)
4e712c1 Daniel Han committed 1 month ago
Diffusion LM training: compiled chunked loss with 0 graph breaks, plus per-family samplers (#13341)
6a70df3 Daniel Han committed 1 month ago
Nemotron-H fast_inference via vLLM weight sharing (fallback on older unsloth_zoo); drop vLLM-only kwargs on fallback (#13440)
8196cf9 Daniel Han committed 1 month ago
Studio MCP: never run server-side tools from the chat tool (#13605)
01e0b8c Daniel Han committed 1 month ago
Bump install.sh / install.ps1 pins to unsloth>=2026.10.5 and unsloth-zoo>=2026.10.5 (#13627)
4bbaf06 Daniel Han committed 1 month ago
Studio: changing the thinking setting also drops a hand Expand (#13625)
781a1a0 Michael Han committed 1 month ago
Studio: emit --moe-cache-mib auto only where the cache measured faster, price the spill micro-batch (#13600)
133a807 Daniel Han committed 1 month ago
Studio tests: restore stub, titlebar backdrop, and run the autoopt parity fixture with diffusers (#13608)
308669b Daniel Han committed 1 month ago
Studio: fix the audio model selector's list, labels and search (#13517)
76b4e2f Etherl committed 1 month ago
Studio: set thinking traces apart from the answer and give tool rows kind icons (#13545)
fa168bb Michael Han committed 1 month ago