Introducing Unsloth Studio!

[shimmyshimmer]

Hey guys, we're super excited to launch Unsloth Studio, a new open-source web UI to train and run LLMs.

Blog + everything you need to know: https://unsloth.ai/docs/new/studio

• Run models locally on Mac, Windows, Linux
• Compare and battle models side-by-side
• Train 500+ models 2x faster with 70% less VRAM
• Supports GGUF, vision, audio, embedding models
• Self-healing Tool calling / web search + code execution
• Auto-create datasets from PDF, CSV, DOCX
• Export models to GGUF, safetensor and more formats

What's Changed

• Update CODEOWNERS for studio and cli by @danielhanchen in Update CODEOWNERS for studio and cli #4266
• [Feature] Support Sequence Classification by @danielhanchen in [Feature] Support Sequence Classification #4264
• [Feature] VLMs support for GRPO by @danielhanchen in [Feature] VLMs support for GRPO #4265
• [Fix] Respect llm_int8_skip_modules for VLM by @danielhanchen in [Fix] Respect llm_int8_skip_modules for VLM #4249
• ROCM support by @danielhanchen in ROCM support #4271
• Remove Blackwell flex attention disable workaround from studio by @danielhanchen in Remove Blackwell flex attention disable workaround from studio #4273
• ROCM support by @danielhanchen in ROCM support #4272
• fix: prevent ai-assist model config RCE via untrusted Hugging Face repos by @danielhanchen in fix: prevent ai-assist model config RCE via untrusted Hugging Face repos #4274
• fix(seed): disable remote code execution in seed inspect dataset loads by @danielhanchen in fix(seed): disable remote code execution in seed inspect dataset loads #4275
• Update CODEOWNERS by @danielhanchen in Update CODEOWNERS #4279
• fix: install data-designer plugin non-editable for Colab compatibility by @LeoBorcherding in fix: install data-designer plugin non-editable for Colab compatibility #4268
• Arch/mixtral by @danielhanchen in Arch/mixtral #4283
• Improve documentation on how to export model from Colab by @danielhanchen in Improve documentation on how to export model from Colab #4284
• feat: Add Mixtral model support by @danielhanchen in feat: Add Mixtral model support #4285
• Initial changes: Refactor Attention by @danielhanchen in Initial changes: Refactor Attention #4286
• patch vlm trainer to resize images by @danielhanchen in patch vlm trainer to resize images #4287
• [WIP] add support for mixtral by @danielhanchen in [WIP] add support for mixtral #4288
• studio: speed up setup -- uv for installs (8x), Ninja for llama.cpp (1.7x) by @danielhanchen in studio: speed up setup -- uv for installs (8x), Ninja for llama.cpp (1.7x) #4289
• fix: remove old comments by @Shine1i in fix: remove old comments #4292
• PR: Windows Setup Improvements by @rolandtannous in PR: Windows Setup Improvements #4299
• miscallenous studio by @Shine1i in miscallenous studio #4293
• Fix: Compare Mode Deadlock, Cancel Event Poisoning & IPC Optimization by @rolandtannous in Fix: Compare Mode Deadlock, Cancel Event Poisoning & IPC Optimization #4303
• studio: fix GGUF inference -- reasoning tokens, max_tokens, server flags, GPU allocation by @danielhanchen in studio: fix GGUF inference -- reasoning tokens, max_tokens, server flags, GPU allocation #4290
• chat only with gguf for mac devices by @Manan17 in chat only with gguf for mac devices #4300
• studio: add max steps and epochs toggle switch by @Imagineer99 in studio: add max steps and epochs toggle switch #4296
• Fix/colab plugin editable install by @LeoBorcherding in Fix/colab plugin editable install #4281
• Graceful shutdown on Windows (signal handlers for Ctrl+C) by @rolandtannous in Graceful shutdown on Windows (signal handlers for Ctrl+C) #4306
• studio: simplify auth UX to password-only login by @Imagineer99 in studio: simplify auth UX to password-only login #4305
• studio: preserve save_steps when toggling to epochs mode by @Imagineer99 in studio: preserve save_steps when toggling to epochs mode #4308
• Fix studio frontend build producing empty Tailwind CSS by @danielhanchen in Fix studio frontend build producing empty Tailwind CSS #4311
• Fix setup.sh crash on Mac with empty gitignore array by @danielhanchen in Fix setup.sh crash on Mac with empty gitignore array #4313
• [Feature] studio: user can upload eval dataset by @Manan17 in [Feature] studio: user can upload eval dataset #4307
• fix: Ctrl+C not terminating backend on Linux by @rolandtannous in fix: Ctrl+C not terminating backend on Linux #4316
• Add download progress bar for non-GGUF models in Chat by @danielhanchen in Add download progress bar for non-GGUF models in Chat #4314
• Apply use_reentrant removal to all TRL trainer configs by @danielhanchen in Apply use_reentrant removal to all TRL trainer configs #4321
• Fix VLM GRPO matmul shape mismatch in _get_per_token_logps_and_entropies by @danielhanchen in Fix VLM GRPO matmul shape mismatch in _get_per_token_logps_and_entropies #4301
• Improve AI Assist: Update default model, model output parsing, logging, and dataset mapping UX by @rolandtannous in Improve AI Assist: Update default model, model output parsing, logging, and dataset mapping UX #4323
• studio: per-model inference defaults, GGUF slider fix, reasoning toggle by @danielhanchen in studio: per-model inference defaults, GGUF slider fix, reasoning toggle #4325
• fix: Resolve CUDA toolkit mismatch on multi-CUDA Windows systems by @rolandtannous in fix: Resolve CUDA toolkit mismatch on multi-CUDA Windows systems #4324
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in [pre-commit.ci] pre-commit autoupdate #4332
• Fix/colab comment edits by @LeoBorcherding in Fix/colab comment edits #4317
• fix: add Qwen3.5 version gate in loader dispatch by @danielhanchen in fix: add Qwen3.5 version gate in loader dispatch #4335
• Fix xformers Blackwell guard: broader coverage and root cause docs by @danielhanchen in Fix xformers Blackwell guard: broader coverage and root cause docs #4338
• studio: improve Colab notebook, redesign ready popup, and clean up install output by @LeoBorcherding in studio: improve Colab notebook, redesign ready popup, and clean up install output #4339
• Add check to disable xformers on newer GPUs by @pluesclues in Add check to disable xformers on newer GPUs #4342
• studio: training progress, CUDA lib path, dataset_num_proc fix by @danielhanchen in studio: training progress, CUDA lib path, dataset_num_proc fix #4336
• studio: fix stale GGUF metadata, update helper model, auth improvements by @danielhanchen in studio: fix stale GGUF metadata, update helper model, auth improvements #4346
• studio: show "Off" for repetition penalty = 1 by @danielhanchen in studio: show "Off" for repetition penalty = 1 #4349
• studio: update Creative/Precise presets, show "Off" for disabled samplers by @danielhanchen in studio: update Creative/Precise presets, show "Off" for disabled samplers #4350
• studio: fix slow cancellation of GGUF generation by @danielhanchen in studio: fix slow cancellation of GGUF generation #4352
• Fix: Remove unused warmupToastShown variable (TS6133) by @rolandtannous in Fix: Remove unused warmupToastShown variable (TS6133) #4353
• Studio: SVG preview, fix streaming and model selector bugs by @danielhanchen in Studio: SVG preview, fix streaming and model selector bugs #4354
• fix: comment out debug print statements by @rolandtannous in fix: comment out debug print statements #4357
• fix(llm_assist): disable thinking mode for helper model JSON output by @rolandtannous in fix(llm_assist): disable thinking mode for helper model JSON output #4358
• studio: improve onboarding UX, tooltips, and training defaults by @danielhanchen in studio: improve onboarding UX, tooltips, and training defaults #4355

New Contributors

• @LeoBorcherding made their first contribution in fix: install data-designer plugin non-editable for Colab compatibility #4268
• @Shine1i made their first contribution in fix: remove old comments #4292
• @Manan17 made their first contribution in chat only with gguf for mac devices #4300
• @Imagineer99 made their first contribution in studio: add max steps and epochs toggle switch #4296

Full Changelog: https://github.com/unslothai/unsloth/commits/March-2026

---

This discussion was created from the release Introducing Unsloth Studio!.

[GReturn]

HUZZAHHHH!!!

[PmNz8]

Thanks for releasing this great tool! Any chances you consider adding simple native installer for windows platform in the future?

[shimmyshimmer]

Working on it! An exe too :)

[bobthe144th]

Yes, that would be very cool, just a simple exe, desktop openable, that would be a serious quality of life improvement over juggling a dozen docker containers, especially if this app used docker desktop for sandboxing

[adrianpuiu]

guys, please add mcp tool support so we can use the mcp tools when generating synthetic datasets

[wasimysaid]

You can already do that! Just add a tool config block and connect it to LLM blocks ! You can add mcps there