New Important Updates!

[shimmyshimmer]

Hey guys, it's only been 2 days since our last release, but we’ve got a lot more important updates:

• Inference is now 20–30% faster. Previously, tool-calling and repeat penalty could slow inference below normal speeds. Inference tokens/s should now perform similar to llama-server / llama.cpp.
• Now Auto-detects older or pre-existing models downloaded from LM Studio, Hugging Face, and similar sources.
• Inference token/s speed is now calculated correctly. Previously, tokens/s included startup time, which made the displayed speed look slower than it actually was. It should now reflect 'true' inference speed.
• CPU usage no longer spikes. Previously, inline querier identity changed every render, causing useLiveQuery to resubscribe continuously.
• Unsloth Studio now has a shutdown x button and shuts down properly. Previously, closing it after opening from the desktop icon would not close it properly. Now, launching from the shortcut also opens the terminal, and closing that terminal fully exits Unsloth Studio. If you still have it open from a previous session you can restart your computer or run lsof -i :8888 then kill -9 <PID>.
• Even better tool-calling and websearch with reduced errors.
• Updated documentation with lots of new info on deleting models, uninstalling etc.
• Cleaner, smarter install and setup logging across Windows and Linux. Output is now easier to read with consistent formatting, quieter by default for a smoother experience, and supports richer --verbose diagnostics when you want full technical detail.
  {% endupdate %}

What's Changed

• Bump installer min version to 2026.3.12 by @danielhanchen in Bump installer min version to 2026.3.12 #4600
• Fix Colab Studio launch and setup.ps1 box alignment by @danielhanchen in Fix Colab Studio launch and setup.ps1 box alignment #4601
• Fix Colab huggingface-hub conflict, ensurepip fallback, bump to 2026.3.14 by @danielhanchen in Fix Colab huggingface-hub conflict, ensurepip fallback, bump to 2026.3.14 #4603
• Update README.md by @rolandtannous in Update README.md #4604
• fix: skip flex_attention for models with non-zero attention_dropout by @Abhinavexists in fix: skip flex_attention for models with non-zero attention_dropout #4605
• Fix Colab setup skipping llama.cpp installation by @rolandtannous in Fix Colab setup skipping llama.cpp installation #4618
• fix: show recommended models in search results by @Shine1i in fix: show recommended models in search results #4615
• studio: align Dataset/Parameters/Training cards, fix expandable height, animate LoRA settings by @Imagineer99 in studio: align Dataset/Parameters/Training cards, fix expandable height, animate LoRA settings #4614
• fix: Windows installer fails on _yaml.pyd Access Denied (os error 5) by @Etherll in fix: Windows installer fails on _yaml.pyd Access Denied (os error 5) #4617
• studio: humanize ETA display for long training runs by @RadouaneElhajali in studio: humanize ETA display for long training runs #4608
• fix: add python-json-logger to data-designer-deps by @Shine1i in fix: add python-json-logger to data-designer-deps #4627
• [Studio] Colab fix - Allow install_python_stack to run on Colab by @rolandtannous in [Studio] Colab fix - Allow install_python_stack to run on Colab #4633
• Fix repetition_penalty default causing 24% TPS drop in GGUF inference by @danielhanchen in Fix repetition_penalty default causing 24% TPS drop in GGUF inference #4634
• fix: install.sh Mac Intel compatibility + Studio no-torch support by @danielhanchen in fix: install.sh Mac Intel compatibility + Studio no-torch support #4624
• tests: add no-torch / Intel Mac test suite by @danielhanchen in tests: add no-torch / Intel Mac test suite #4646
• fix: use unsloth[huggingfacenotorch] instead of --no-deps in no-torch mode by @danielhanchen in fix: use unsloth[huggingfacenotorch] instead of --no-deps in no-torch mode #4647
• Fix Gemma3N audio training stride assertion with non-reentrant checkpointing by @danielhanchen in Fix Gemma3N audio training stride assertion with non-reentrant checkpointing #4629
• Fix missing num_items_in_batch in unsloth_prediction_step by @danielhanchen in Fix missing num_items_in_batch in unsloth_prediction_step #4616
• Make Studio shortcuts launch in a visible terminal by @danielhanchen in Make Studio shortcuts launch in a visible terminal #4638
• studio: setup log styling by @Imagineer99 in studio: setup log styling #4494
• Fix ~1.2s TTFT penalty when tools are enabled in Studio by @danielhanchen in Fix ~1.2s TTFT penalty when tools are enabled in Studio #4639
• Fix GGUF GPU fit check to account for KV cache VRAM by @danielhanchen in Fix GGUF GPU fit check to account for KV cache VRAM #4623
• feat: update app icons to rounded logo by @Shine1i in feat: update app icons to rounded logo #4640
• Streaming tool detection: guard late tool_calls, filter incomplete fragments by @danielhanchen in Streaming tool detection: guard late tool_calls, filter incomplete fragments #4648
• fix: install no-torch runtime deps via requirements file by @danielhanchen in fix: install no-torch runtime deps via requirements file #4649
• Fix orphan server cleanup killing user's own llama-server by @danielhanchen in Fix orphan server cleanup killing user's own llama-server #4622
• fix: add auth + UX improvements to shutdown button by @Shine1i in fix: add auth + UX improvements to shutdown button #4642
• Fix inference failing for transformers 5.x models (trust_remote_code) by @danielhanchen in Fix inference failing for transformers 5.x models (trust_remote_code) #4652
• fix: no-torch install deps without pulling torch transitively by @danielhanchen in fix: no-torch install deps without pulling torch transitively #4650
• Detect always-on reasoning models and show Think button as locked-on by @danielhanchen in Detect always-on reasoning models and show Think button as locked-on #4654
• fix: replace navbar shutdown text button with icon-only button by @Shine1i in fix: replace navbar shutdown text button with icon-only button #4655
• Fall back to parsing model name when HF API has no param count by @danielhanchen in Fall back to parsing model name when HF API has no param count #4656
• fix: disable OCR in pymupdf4llm PDF extraction by @Shine1i in fix: disable OCR in pymupdf4llm PDF extraction #4659
• Fix HF cache default and show LM Studio models in chat/inference by @rolandtannous in Fix HF cache default and show LM Studio models in chat/inference #4653
• Bump minimum unsloth version to 2026.3.16 in install scripts by @danielhanchen in Bump minimum unsloth version to 2026.3.16 in install scripts #4663

New Contributors

• @Abhinavexists made their first contribution in fix: skip flex_attention for models with non-zero attention_dropout #4605
• @RadouaneElhajali made their first contribution in studio: humanize ETA display for long training runs #4608

Full Changelog: v0.1.2-beta...v0.1.25-beta

---

This discussion was created from the release New Important Updates!.