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
Unsloth Studio: Cosmos3 256p, NVIDIA's frame ranges and a per-model frame-rate picker (#16737)
681fd01 Leo Borcherding committed 10 days ago
Long-chat gate: Firefox leg for dialogs, menus, browser pane and typing (#17361)
56fa8af Daniel Han committed 10 days ago
Studio diffusion: keep a hosted checkpoint's header across processes, so an 8 GiB Qwen-Image-2.1 load skips its FP8 header read (L4 load -1.8 s) (#17353)
0a9c409 Daniel Han committed 10 days ago
Studio sd.cpp: send the simple scheduler so flow families get the diffusers sigma grid (#17360)
49e0073 Daniel Han committed 10 days ago
Studio diffusion: keep no pinned host copy of a resident Qwen-Image-2.1 text encoder (L4 host RAM -8.1 GiB, peak RSS 25.3 -> 16.6 GiB) (#17359)
b14e78d Daniel Han committed 10 days ago
Studio MiniMax-H3: park the VAEs when a render returns, so later requests keep their residency budget (#17356)
3bcaf2d Daniel Han committed 10 days ago
Studio MiniMax-H3: keep conditioner layers on the card beside a resident denoiser, size its host pins to their weights (#17355)
4d709a5 Daniel Han committed 10 days ago
Studio MiniMax-H3: skip the up-front denoiser slab pin when every block will be resident (-16 GiB host peak on A100 40 GB) (#17354)
61e18dd Daniel Han committed 10 days ago
Studio: fit a GGUF reload only after the stopped llama-server's VRAM shows as free (#17345)
d4c684e oobabooga committed 10 days ago
Studio: open and close dialogs and menus in long chats without a full thread re-layout on Firefox (#17357)
0c46cdb Daniel Han committed 10 days ago
Studio: open a long chat at its latest message on Chromium (#17351)
c74b6cf Daniel Han committed 10 days ago
Studio sd.cpp: retry an image render that sd-server refused for VRAM under half the budget (#17339)
f2e923e oobabooga committed 10 days ago
Studio: compute ROCm VAE half-math SDPA per call instead of flipping the global flag (#17350)
3d4605c Daniel Han committed 10 days ago
Studio: restore ROCm TunableOp state when the last diffusion pipeline unloads (#17340)
e5aeca4 Daniel Han committed 10 days ago
Studio: glass toast surface for the runtime, Studio and Desktop update cards (#17352)
3b315e4 Michael Han committed 10 days ago
Studio sd.cpp: send MiniMax-H3 reference video and audio to the resident sd-server (#17349)
9b789e3 Daniel Han committed 10 days ago
Studio: an unverified cached .pt no longer demotes the baseline-format safetensors twin (#17342)
f6dffef Daniel Han committed 10 days ago
Studio ROCm layer split: apply the common SDPA guard only once the split fits (#17347)
2d0774e Daniel Han committed 10 days ago
Studio sd.cpp: only block-split on builds whose --auto-fit takes on|off (#17346)
4e92937 Daniel Han committed 10 days ago
Studio diffusion: do not let functools.wraps copy the prompt-cache transparent-wrapper marker (#17348)
cef96a9 Daniel Han committed 10 days ago
Studio FLUX.3: probe strided cuDNN attention on every layer split card (#17341)
e12eb2e Daniel Han committed 10 days ago
Studio: reroute only official NVFP4 checkpoints to INT8 on pre-Blackwell GPUs (#17343)
5aa8989 Daniel Han committed 10 days ago
Studio: share the FLUX.3 joint block int8 ConvRot quant under the span forward (#17337)
762e849 Daniel Han committed 10 days ago
Long-chat gate: block on CPU / paint growth with chat length, judged on fresh-page rechecks (#17336)
ff80d4e Daniel Han committed 10 days ago
Studio: open the browser pane without React's Suspense reveal hold (#17335)
ec6d160 Daniel Han committed 10 days ago
Studio ROCm RDNA int8: Triton W8A8 kernels and fused Qwen-Image-2.1 projections (gfx1151 INT8 -12% per step, explicit int8 -26% to -34%) (#17334)
57c9573 Daniel Han committed 10 days ago
Studio CI: WebKit leg for the long-chat gate, report-only paint / CPU length check, recheck gap fixed (#17331)
3fe5ade Daniel Han committed 10 days ago
Fix Windows wallpaper titlebar and Transparent glass surfaces (#16959)
d071aff Lee Jackson committed 10 days ago
Unsloth Studio: restyle a video with Cosmos3-Nano-Sim-Transfer on sd-server (#16975)
945d2b6 Leo Borcherding committed 10 days ago
Studio: WebKitGTK Glass drops the sidebar's live blur and Frosted's grain tile (long-chat scroll) (#17327)
c96493c Daniel Han committed 10 days ago