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: stop refusing a diffusion load on a DGX Spark against its own download (#10705)
0873389 Daniel Han committed 2 months ago
Cover the restore of every variable the install.ps1 handoff block saves (#10566)
2bd5dab Daniel Han committed 2 months ago
Reduce comment volume across the installers, CLI, unsloth package and backend utilities (#10504)
11bed73 Daniel Han committed 2 months ago
Reduce comment volume across the studio frontend and backend routes (#10503)
8e31299 Daniel Han committed 2 months ago
Studio: keep a chat working after switching to a model without tool support (#10454)
8d48887 Nilay committed 2 months ago
fix: exclude padded logits from softcapped cross entropy (#10616)
57b1d75 Chenghao Liu committed 2 months ago
fix: materialize strided gradients before RMSNorm backward (#10617)
8a37435 Chenghao Liu committed 2 months ago
Studio: recover the GPU capacity nvidia-smi will not report on a DGX Spark (#10703)
0392cb2 Daniel Han committed 2 months ago
Studio: fit the composer on narrow screens, and scale the greeting with it (#10694)
4c357a1 Michael Han committed 2 months ago
Studio: keep the output a timed-out python or terminal call already produced (#10664)
d7bc72c Nilay committed 2 months ago
Windows No Compiler CI: see the intermediates the compiler cleaned up (#10711)
a167f28 Daniel Han committed 2 months ago
CI: trigger Studio smokes on what they observe, run install matrices as a PR subset, fold Chat UI shards (#10689)
9e8b747 Daniel Han committed 2 months ago
Studio: keep an uploaded document's own name instead of replacing it with underscores (#10667)
89b7ff0 Nilay committed 2 months ago
Studio: keep literal <think> text visible when the request turns thinking off (#10662)
1ceffa7 Nilay committed 2 months ago
Studio: fail a Deep Research run when no step gathered any evidence (#10663)
7ead40f Nilay committed 2 months ago
Block markerless execution-class tool-call promotion in Studio (#10507)
30584ef Daniel Han committed 2 months ago
Studio: stop pricing an Anthropic image by its base64 length in KV admission (#10669)
db3e48a Nilay committed 2 months ago
Studio: gate GGML_CUDA_P2P on a confirmed NVLink fabric, not a product name (#10646)
10a95da Daniel Han committed 2 months ago
Raise the startup bundle budget to where main actually is (#10702)
b50d53d Daniel Han committed 2 months ago
Read the installers' own code, not the scripts they emit (#10701)
bd8df9b Daniel Han committed 2 months ago
Studio: only strip a valid trailing __IMAGES__/__RAG_SOURCES__ envelope (#10668)
d4ef69f Nilay committed 2 months ago
Studio: stop a generation recovery deleting the turn's tool calls (#10584)
7436c10 Michael Han committed 2 months ago
Studio: keep the LoRA rank, alpha and variant when leaving CPT (#10555)
7850618 Nilay committed 2 months ago
Unbreak main: compare the autoload guard's conditions, not its exact text (#10654)
92141b3 Daniel Han committed 2 months ago
Studio: reuse the named CLI API key across studio run (#10608)
1b0a2e4 Chaitanya Laxman committed 2 months ago
Fold duplicated backend, hub and package tests into parametrized runs (#10499)
38f78af Daniel Han committed 2 months ago
Fix Deep Research JSON output on MLX and Transformers (#10611)
d9caa7c Lee Jackson committed 2 months ago