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
Fused QKNorm + RoPE: keep type promotion for mixed dtype tables, reuse grad in the trainable norm fallback (#14200)
1fe718b Daniel Han committed 1 month ago
GRPO: file compute_loss metrics by model.training, not control.should_evaluate (#14112)
e0b8879 Vineeth Sai Varikuntla committed 1 month ago
Studio: fix browser toolbar alignment and narrow layouts (#14018)
e469bfe Lee Jackson committed 1 month ago
Fix padded prefill sliding window being one key too wide (#13906)
063c703 Vineeth Sai Varikuntla committed 1 month ago
Studio: start the FLUX.3 prompt upsampler while Enhance prompt is on (#14208)
a8cf93a Daniel Han committed 1 month ago
Studio: follow a same-model MLX reload in the KV width and int8 prefill settings (#14210)
06223a9 Daniel Han committed 1 month ago
Copy the residual before the in-place norm in Granite fast decode (#13454)
e7252ff Vineeth Sai Varikuntla committed 1 month ago
Studio: compile FLUX.3 Klein once for every prompt length (#14211)
4d353f5 Daniel Han committed 1 month ago
Hybrid packing + block compile: fix the guard crash, the skipped varlen, and the cross-document leak (#13956)
a33463f Etherl committed 1 month ago
Studio: raise the startup bundle budget for #14197 (#14203)
0174c53 Daniel Han committed 1 month ago
Studio: strip tool markup from streamed text without rescanning the whole turn per token (#14143)
1de12ab Daniel Han committed 1 month ago
Studio: keep the FLUX.3 Image Klein GGUF row under its Fast FP8 row at the top of the Images picker (#14197)
d5e700b Daniel Han committed 1 month ago
Studio: raise the startup bundle budget for #14190 (#14198)
94464ba Daniel Han committed 1 month ago
Studio: open //host links as web links; run apt-get by absolute path; ignore the dev Studio home (#14190)
551dc28 Daniel Han committed 1 month ago
Studio: managed accounts use the server's own HF endpoint for recipe seeds (#14187)
e42df82 Daniel Han committed 1 month ago
Notebooks CI: refresh the Colab oracle snapshots to the current GPU runtime (#14182)
4dc8bf0 Daniel Han committed 1 month ago
Studio: update Lemonade to 2026.42.1 and FastFlowLM to v1.0.9 (#14189)
f2ebbb7 Daniel Han committed 1 month ago
Studio: keep the prompt upsampler client out of the startup chunks, reset the startup budget to the build (#14168)
69dbf9e Daniel Han committed 1 month ago
Desktop: bump rustls to 0.23.46 and quinn-proto to 0.11.19 for RustSec advisories (#14186)
6b60d58 Daniel Han committed 1 month ago
Studio: pin FLUX.3 Image Klein (Fast FP8) to the top of the Images picker (#14184)
b227654 Daniel Han committed 1 month ago
Studio: one tooltip design everywhere, matching chat's text colours in light and dark (#14172)
8bb21cc Michael Han committed 1 month ago
Studio: run the Windows sandbox (MXC) on ARM64 PCs with Microsoft's ARM64 build (#14176)
e4ff174 Daniel Han committed 1 month ago
CLI: doctor --mcp sends the API key only to a Studio that proves its identity (#14178)
0c23763 Daniel Han committed 1 month ago
Studio: darker unselected model picker toggle, collapse icon for Compact conversation (#14183)
aacd821 Michael Han committed 1 month ago
Desktop: keep DMA-BUF off on NVIDIA Wayland when compositing is turned off by the environment (#14141)
f4b9af0 Daniel Han committed 1 month ago
Studio: match letter shortcuts by the letter the layout types (#14113)
78e5eef Sourav Rajvi committed 1 month ago
Studio: label the Chat with files composer pill Files instead of RAG (#14179)
2fee36f Michael Han committed 1 month ago
Studio: download the UD-Q4_K_XL FLUX.3 prompt upsampler by default (#14177)
8675c63 Daniel Han committed 1 month ago