Unsloth

Hey guys, we're doing one more-ish update before a major revamp which is likely coming this week or next week. Our revamp will change a lot of things, especially with new major features and a lot of design changes.

• NEW: API calling support now with image generation + editing, proper web search, code execution, auto prompt caching. Connect OpenAI, Anthropic and more.
• unsloth studio update should now work properly. For Mac users, use the install curl command instead: curl -fsSL https://unsloth.ai/install.sh | sh
• Proper support for non-English languages e.g. Japanese, Chinese, Indian etc.

https://github.com/user-attachments/assets/70456395-e016-4273-8256-35adb206267e

Many of you may have missed our previous release which only lasted for one day. We introduced:

• Connect to external inference backends: vLLM, Ollama, llama-server
• Security improvements

• Auto MTP speculative decoding for MTP GGUFs; get the best settings customized for your hardware.

API provider calling & external connections

• You can now connect Unsloth to any API cloud provider (OpenAI, Anthropic, OpenRouter etc.)
• Built-in web search for OpenAI, Anthropic, OpenRouter and Kimi
• Built-in code execution for OpenAI and Anthropic (Anthropic containers persist and are reused across turns)
• Prompt caching is enabled for OpenAI and Anthropic models saving 50 to 90% of costs.
• Image generation + editing
• API key now optional for local providers (llama.cpp / vLLM / Ollama)
• Auto-load models when adding a cloud provider

Other Unsloth Studio updates

• OpenDocument chat attachments
• o3 reasoning summary payload
• Sending/prompting non-English languages (e.g. Japanese, Chinese) now works properly
• IME composer hardening, RTL dir="auto", long log-line truncation fix
• Tool reasoning trace rendering in UI
• Fully offline support: cached GGUF discovery and offline DNS auto-detect for both inference and training

Unsloth Studio security improvements

• Authentication rate-limiting, proxy-aware so reverse proxies don't bypass it
• Sandboxed worker with a tightened blocklist (bash, hf upload, NOFILE)
• Path containment so workers can't escape their in-flight tmp dirs
• Strict schema validation across the Studio API
• Tightened CSP / security headers (only legitimate favicon hosts allowed)
• Removed the torch.load fallback on training_args.bin so untrusted pickles can never execute on model load
• Hardened Tauri desktop release flow
• Frontend auth: singleflight token refresh, current-password input on changes, working logout, shared 422 helper
• Cancel cleanup now scoped strictly to in-flight tmp dirs so it can never delete user state

What's Changed

• install: bump unsloth floor to >=2026.5.5 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5621
• Studio: persist chat toggles and preserve custom sampling by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5587
• Studio: add connections toggle and order hosted providers by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5588
• studio/ci: harden three pre-existing CI flakes by @danielhanchen in https://github.com/unslothai/unsloth/pull/5627
• Studio: expand Connections model picker for local inference server by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5643
• Move uninstall scripts into scripts/ and fix references by @danielhanchen in https://github.com/unslothai/unsloth/pull/5644
• Studio: provider model loading controls by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5645
• Studio: unify connection copy by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5654
• Remove dead chat suggestions wiring by @alkinun in https://github.com/unslothai/unsloth/pull/5665
• Studio: Claude Code Anthropic API tool compatibility by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5390
• studio/frontend: friendlier 404 fallback for unknown routes by @danielhanchen in https://github.com/unslothai/unsloth/pull/5664
• chore(deps): bump the npm-oxc-validator group across 1 directory with 2 updates by @dependabot[bot] in https://github.com/unslothai/unsloth/pull/5667
• studio/chat: Disable send button for empty composer by @harryfrzz in https://github.com/unslothai/unsloth/pull/5647
• studio/frontend: correct Think pill aria-label before model loads by @danielhanchen in https://github.com/unslothai/unsloth/pull/5655
• studio/frontend: fix onboarding CSP violations by @danielhanchen in https://github.com/unslothai/unsloth/pull/5658
• studio/frontend: set per-route document.title by @danielhanchen in https://github.com/unslothai/unsloth/pull/5660
• Fix Windows Tauri build and signing by @wasimysaid in https://github.com/unslothai/unsloth/pull/5694
• Respect GC for GRPO by @Datta0 in https://github.com/unslothai/unsloth/pull/5269
• studio: unblock /load event loop on detect_audio_type (#5642, #5635) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5669
• studio: settle GPU VRAM after killing llama-server before the next reload by @danielhanchen in https://github.com/unslothai/unsloth/pull/5693
• Studio: surface prompt-cache token counts in /v1/chat/completions usage chunk by @danielhanchen in https://github.com/unslothai/unsloth/pull/5670
• Studio: per-model Anthropic server-side tool versions by @danielhanchen in https://github.com/unslothai/unsloth/pull/5679
• Studio: support Anthropic 1h cache TTL via prompt_cache_ttl by @danielhanchen in https://github.com/unslothai/unsloth/pull/5685
• Studio: wire OpenAI image_generation tool by @danielhanchen in https://github.com/unslothai/unsloth/pull/5688
• Studio: per-session cost calculator + /api/providers/pricing endpoint by @danielhanchen in https://github.com/unslothai/unsloth/pull/5690
• Studio: wire Anthropic web_fetch server-side tool by @danielhanchen in https://github.com/unslothai/unsloth/pull/5671
• Studio: persist chat history in backend storage by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5272
• Studio: wire Anthropic server-side context compaction by @danielhanchen in https://github.com/unslothai/unsloth/pull/5686
• Studio: wire OpenAI Responses server-side context compaction by @danielhanchen in https://github.com/unslothai/unsloth/pull/5687
• Studio: PDF / document attachments for Anthropic + OpenAI by @danielhanchen in https://github.com/unslothai/unsloth/pull/5689
• Studio: reconcile external providers across browsers after delete by @danielhanchen in https://github.com/unslothai/unsloth/pull/5698
• Studio: persist external provider selection across page refresh by @danielhanchen in https://github.com/unslothai/unsloth/pull/5697
• Studio: add Anthropic and OpenAI prompt guards for disabled tools by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5674
• Studio: persist external checkpoint when picker uses setParams by @danielhanchen in https://github.com/unslothai/unsloth/pull/5700
• Studio: surface OpenAI image_generation as composer Images pill by @danielhanchen in https://github.com/unslothai/unsloth/pull/5699
• Studio: expose Anthropic 5m vs 1h prompt cache TTL in Configuration by @danielhanchen in https://github.com/unslothai/unsloth/pull/5703
• Fix connected chat model selection after refresh by @wasimysaid in https://github.com/unslothai/unsloth/pull/5702
• Studio: render OpenAI image_generation results inline in chat by @danielhanchen in https://github.com/unslothai/unsloth/pull/5705
• Truncate long code execution tool output by @wasimysaid in https://github.com/unslothai/unsloth/pull/5708
• fix(gpt-oss): prefer flex attention over sdpa by @Datta0 in https://github.com/unslothai/unsloth/pull/5701
• Bump install.sh / install.ps1 pin to unsloth>=2026.5.6 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5716
• ci: unblock Studio Windows + Linux + Mac smoke (supersedes #5733, #5734, #5738) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5741
• ci: broaden Linux + narrow Windows llama.cpp runtime patterns + trim #5741 comments by @danielhanchen in https://github.com/unslothai/unsloth/pull/5746
• Lower default RL weight_decay from 0.01 to 0.001 for LoRA by @danielhanchen in https://github.com/unslothai/unsloth/pull/5747
• Studio: strip orphan tool_call XML leaking into visible content by @danielhanchen in https://github.com/unslothai/unsloth/pull/5735
• Bump install.sh / install.ps1 pin to unsloth>=2026.5.7 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5753
• Fix MLX Studio base model export save method by @Lyxot in https://github.com/unslothai/unsloth/pull/5727
• fix(chat_templates): check find() return value before slicing on placeholders by @Ricardo-M-L in https://github.com/unslothai/unsloth/pull/5763
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in https://github.com/unslothai/unsloth/pull/5773
• Studio: stop seeded admin to cross-origin callers by @danielhanchen in https://github.com/unslothai/unsloth/pull/5739
• Studio: surface external-provider cache hits and writes in context bar by @danielhanchen in https://github.com/unslothai/unsloth/pull/5736
• Studio: Anthropic fast_mode toggle and streaming refusal handling by @danielhanchen in https://github.com/unslothai/unsloth/pull/5715
• Studio: rewrite OpenAI Responses citation markers to markdown links by @danielhanchen in https://github.com/unslothai/unsloth/pull/5713
• Studio: standalone Fetch pill for Anthropic web_fetch by @danielhanchen in https://github.com/unslothai/unsloth/pull/5742
• Studio: surface Anthropic document citations inline + in Sources panel by @danielhanchen in https://github.com/unslothai/unsloth/pull/5718
• Studio: pricing follow-up to #5690 (longest-prefix match + chat-style usage keys) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5722
• Support follow-up edits for generated images by @wasimysaid in https://github.com/unslothai/unsloth/pull/5712
• Recipe Studio local model selector by @wasimysaid in https://github.com/unslothai/unsloth/pull/5769
• Improve image generation UI by @wasimysaid in https://github.com/unslothai/unsloth/pull/5784
• Studio: per-card web_search result + shell_call output fallback (OpenAI) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5785
• Keep generated image loading dots animated by @wasimysaid in https://github.com/unslothai/unsloth/pull/5786
• Studio: auto-recover when shadowed 'unsloth' on PATH hides the frontend dist by @danielhanchen in https://github.com/unslothai/unsloth/pull/5782
• studio/frontend: show "1 second" instead of "1 seconds" in thinking blocks by @oobabooga in https://github.com/unslothai/unsloth/pull/5777
• Studio: stop the model from replying twice when it refuses by @oobabooga in https://github.com/unslothai/unsloth/pull/5775
• studio/frontend: add space above the scroll-to-bottom arrow in chat by @oobabooga in https://github.com/unslothai/unsloth/pull/5776
• Fix unsloth studio update silently downgrading on macOS arm64 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5767

New Contributors

• @harryfrzz made their first contribution in https://github.com/unslothai/unsloth/pull/5647
• @Lyxot made their first contribution in https://github.com/unslothai/unsloth/pull/5727
• @oobabooga made their first contribution in https://github.com/unslothai/unsloth/pull/5777

Full Changelog: https://github.com/unslothai/unsloth/compare/v0.1.41-beta...0.1.417-beta