v0.1.405-beta

We've got lots of new updates. Please use the latest Unsloth `v0.1.405-beta`, not `v0.1.40-beta` which is older.
* **~2x faster GGUF inference** with automatically enabled MTP
* [API calling support](https://unsloth.ai/docs/integrations/connections) for [OpenAI](https://unsloth.ai/docs/integrations/connections/openai), [Anthropic](https://unsloth.ai/docs/integrations/connections/anthropic-claude) etc. with auto prompt caching, web search, code execution
* Connect to external inference backends: [vLLM](https://unsloth.ai/docs/integrations/connections/vllm), [Ollama](https://unsloth.ai/docs/integrations/connections/ollama), [llama-server](https://unsloth.ai/docs/integrations/connections/connect-llama.cpp-to-unsloth-run-ggufs-with-llama-server)
* Experimental **MLX inference**
* Proper support for **non-English languages**
* **Security improvements**

## MTP speculative decoding support 1.4 to 2x faster inference!
- **Auto MTP speculative decoding** for MTP GGUFs; warn when the bundled llama.cpp prebuilt is stale or too old for MTP
- New pre-built llama.cpp binaries for MTP support!

## 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.
- API key now optional for local providers (llama.cpp / vLLM / Ollama)
- Auto-load models when adding a cloud provider

## MLX inference (Experimental)
- MLX quants and models now can run locally on your Mac machines!
- We'll be adding thinking, tools and web search soon!

## 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
- Lots of UI/UX polish: dark theme refactor, right sidebar redesign, time-of-day sloth mascot, dismissable copyable toasts, larger chat composer, code-execution config polish, composer action pill styling, narrower Discord button

## Training updates
- Gemma attention mask fixes
- Multi Image GRPO
- GRPO hidden-state return experiments
- New Continued Pretraining (CPT) training method as a first-class option
- Gemma-4 MoE LoRA extractor registered to fix `grouped_mm` contraction crash
- Opt-in fused `lm_head` + cross-entropy forward, with single-matmul path under `UNSLOTH_RETURN_LOGITS=1`
- Pass batch size for eval
- Eval/training paths now honour `HF_DATASETS_OFFLINE` alongside `HF_HUB_OFFLINE`

## 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

## Bug fixes and correctness
- Layout-aware MoE LoRA merge with loud-fail on fallback (no more silent wrong saves)
- `num_logits_to_keep` regression fixed on transformers >= 4.52
- Preserve tokenizer EOS token on merged saves
- Resume PEFT checkpoints under sentence-transformers >= 5.4
- Restore Flash > SDPA > Flex attention priority for non-Gemma3 models
- ORPO text-only tokenization now works with processors
- Embedding matrix size mismatch fix
- Vicuna chat template fix
- `fast_generate` unifies legacy and new logits kwargs (fixes Mistral merge site)
- `higher_precision_softmax` made idempotent
- Patch every `LOSS_MAPPING` key aliased to `ForCausalLMLoss` (covers transformers 5.x)
- GGUF converter sibling imports fixed
- UTF-8 encoding added to all text-mode file operations
- Serialise GGUF reload and inherit `unsloth-run` extra args
- Fix `/recommended-folders` 500 on unreadable model directories under Python 3.12+
- Cross-family GGUF projector blocked in flat local dirs (no more wrong-vision-tower loads)

## Installer and platform reliability
- Custom install paths via `STUDIO_HOME` / `UNSLOTH_STUDIO_HOME`
- CPU-only Linux x86_64 routed to `ggml-org/llama.cpp` prebuilts
- Windows CUDA install fixes: paired `cudart` bundle and Torch NVIDIA DLL paths added to `PATH`
- Skip `flash-attn` install on Blackwell GPUs (sm_100+)
- Refresh Intel XPU extras for torch 2.7.1 / 2.9.1 / 2.10 / 2.11.0 / 2.12.0; torch upper cap raised to <2.13.0
- HIP source builds on Ubuntu 24.04 now inject `--gcc-install-dir`
- Linux prebuilt fixes for branch-based llama.cpp releases (mangled symlink repair, top-level dir strip)
- New uninstallers for Linux, macOS (`uninstall.sh`) and Windows (`uninstall.ps1`)
- Mac desktop shortcut spawning and lifecycle fixed
- `unsloth --version` flag
- Studio web update banner and release version display
- GPU pinned at 95% headroom, with a warning on silent CPU fallback
- Auto-install flash-linear-attention and tilelang for Qwen3.5 family

## What's Changed in Unsloth
* Bump installer floor to 2026.5.2 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5297
* install: support STUDIO_HOME / UNSLOTH_STUDIO_HOME for custom install paths by @danielhanchen in https://github.com/unslothai/unsloth/pull/5190
* Route CPU-only Linux x86_64 to ggml-org/llama.cpp prebuilts by @danielhanchen in https://github.com/unslothai/unsloth/pull/5302
* feat(studio): MLX training tab on Apple Silicon (LoRA / full FT, VLM, export) by @Manan17 in https://github.com/unslothai/unsloth/pull/5265
* feat(studio): add Continued Pretraining (CPT) as a training method by @OnePunchMonk in https://github.com/unslothai/unsloth/pull/4677
* Fix 14 stale tests under tests/studio/install/ that drifted from code by @danielhanchen in https://github.com/unslothai/unsloth/pull/5305
* Add Studio PR-time CI: pin enforcement, frontend, backend, wheel smoke by @danielhanchen in https://github.com/unslothai/unsloth/pull/5298
* Studio: restore Studio API and Help menu UI by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5310
* [studio]: Fix tool reasoning trace in UI  by @CodeMan62 in https://github.com/unslothai/unsloth/pull/5314
* fix: 3 patch_* helpers — fast_lora import, sft_trainer Union, openenv OSError by @danielhanchen in https://github.com/unslothai/unsloth/pull/5319
* Studio: API settings overflow with long Colab URLs by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5286
* tests/studio/install: parallel UNSLOTH_STUDIO_HOME smoke test by @danielhanchen in https://github.com/unslothai/unsloth/pull/5306
* Studio: Dark theme refactor, right sidebar redesign, and chat UI polish by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5150
* fix: harden Studio IME composer sends by @Etherll in https://github.com/unslothai/unsloth/pull/5327
* Studio: stop truncating long log lines as suspected base64 by @rolandtannous in https://github.com/unslothai/unsloth/pull/5335
* fix(gh_client): fail fast on 401/403 auth errors instead of retrying forever (#5325) by @Anai-Guo in https://github.com/unslothai/unsloth/pull/5329
* fix: unblock 4 tests deselected/skipped in #5312 (real bugs) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5359
* fix(tests/sh): accept pinned tokenizers line after #5359 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5361
* CI: scope GITHUB_TOKEN permissions, add MLX CI, unblock ~60 skipped tests by @danielhanchen in https://github.com/unslothai/unsloth/pull/5312
* studio/tests: make Playwright model-selector probe best-effort by @danielhanchen in https://github.com/unslothai/unsloth/pull/5371
* Studio: download paired cudart bundle on Windows CUDA installs by @danielhanchen in https://github.com/unslothai/unsloth/pull/5322
* Studio: add torch's pip nvidia DLL dirs to PATH on Windows by @danielhanchen in https://github.com/unslothai/unsloth/pull/5324
* studio: authenticate HF downloads across Studio CI workflows by @danielhanchen in https://github.com/unslothai/unsloth/pull/5370
* dependabot: group security updates and cover /studio/frontend npm advisories by @danielhanchen in https://github.com/unslothai/unsloth/pull/5372
* Add Studio web update banner and release version display by @wasimysaid in https://github.com/unslothai/unsloth/pull/5308
* ci/install: retry transient github.com 5xx on unsloth-zoo git fetches by @danielhanchen in https://github.com/unslothai/unsloth/pull/5389
* studio/ci: pre-install lockfile supply-chain audit (npm + cargo) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5392
* studio/ci: npm tarball content scanner (no-install, hostile-input safe) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5393
* studio/tests: AbortSignal-bound in-page fetches and wall-clock watchdog for Playwright probes by @danielhanchen in https://github.com/unslothai/unsloth/pull/5391
* chore: remove unused .semgrep/unsloth-rules.yml by @danielhanchen in https://github.com/unslothai/unsloth/pull/5395
* studio/ci: sweep actions/cache v5 hardening across sibling smoke workflows by @danielhanchen in https://github.com/unslothai/unsloth/pull/5399
* studio/ci: harden HF_HOME cache against actions/cache v5 silent restore failures by @danielhanchen in https://github.com/unslothai/unsloth/pull/5396
* Harden Tauri release flow by @wasimysaid in https://github.com/unslothai/unsloth/pull/5341
* Gemma attn by @Datta0 in https://github.com/unslothai/unsloth/pull/5346
* Multi Image GRPO by @Datta0 in https://github.com/unslothai/unsloth/pull/5197
* [GRPO] Try returning hidden statex for GRPO by @Datta0 in https://github.com/unslothai/unsloth/pull/5142
* Studio: pin GPU at 95% headroom and warn on silent CPU fallback by @danielhanchen in https://github.com/unslothai/unsloth/pull/5323
* Chore(deps): bump the actions group across 1 directory with 4 updates by @dependabot[bot] in https://github.com/unslothai/unsloth/pull/5394
* security: NOT affected by Mini Shai-Hulud (May-12 wave) -- forward-looking hardening only by @danielhanchen in https://github.com/unslothai/unsloth/pull/5397
* studio: security and hardening pass (auth rate-limit, sandbox, path containment, schema validation, headers) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5375
* studio: fix training page regressions from the security hardening pass by @rolandtannous in https://github.com/unslothai/unsloth/pull/5409
* Studio: parity of thinking trace icon with Think toggle icon by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5407
* Studio: vary empty chat sloth mascot by local time of day by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5354
* security: persist-credentials:false on every actions/checkout (org-wide sweep) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5413
* import_fixes: stub transformers.conversion_mapping so peft 0.19.x imports on transformers 4.x by @danielhanchen in https://github.com/unslothai/unsloth/pull/5416
* chore: trim verbose comments added in PR #5416 (commit 12295c1f) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5418
* studio/ci: flat GGUF+mmproj cache for Mac json-images smoke, save partial caches on cancel by @danielhanchen in https://github.com/unslothai/unsloth/pull/5417
* studio: comment out training_args.bin torch.load fallback in model_config by @danielhanchen in https://github.com/unslothai/unsloth/pull/5419
* tests: import_fixes drift detectors (HARD GATE on Core matrix) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5414
* tests: drift detector parity with unsloth-zoo (fix Core matrix RED on triton + vllm) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5421
* scripts: ship deterministic comment / docstring-only diff verifier by @danielhanchen in https://github.com/unslothai/unsloth/pull/5422
* studio: API external provider support for chat (OpenAI, Mistral, Gemini, Cohere, Anthropic, OpenRouter, DeepSeek, custom providers) by @rolandtannous in https://github.com/unslothai/unsloth/pull/4706
* import_fixes + drift detectors: cover transformers 5.x drift (unblocks PR #5376) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5423
* MLX training support for Studio on Apple Silicon by @mmathew23 in https://github.com/unslothai/unsloth/pull/5340
* studio: drop unused max_grad_value schema + route plumbing by @danielhanchen in https://github.com/unslothai/unsloth/pull/5424
* Studio: Passing batch size for eval by @uderbashi in https://github.com/unslothai/unsloth/pull/5168
* studio: skip flash-attn install on Blackwell GPUs (sm_100+) by @rolandtannous in https://github.com/unslothai/unsloth/pull/5420
* Fix: Add missing utf-8 encoding to text-mode file operations by @Tenith01 in https://github.com/unslothai/unsloth/pull/5356
* Fix/issue 3667 vicuna template by @Tenith01 in https://github.com/unslothai/unsloth/pull/5357
* tests: public-api surface drift detector (companion to test_import_fixes_drift.py) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5428
* add UNSLOTH_ALLOW_CPU=1 path for CPU-only CI / source-inspection tests by @danielhanchen in https://github.com/unslothai/unsloth/pull/5429
* fix(studio/mmproj): block cross-family projectors in flat local GGUF dirs (#5347) by @Anai-Guo in https://github.com/unslothai/unsloth/pull/5350
* studio/mmproj: skip unwanted GGUF values via seek instead of read by @danielhanchen in https://github.com/unslothai/unsloth/pull/5431
* ci: install ipython so transformers.utils.notebook imports cleanly in zoo pytest by @danielhanchen in https://github.com/unslothai/unsloth/pull/5437
* studio/mlx: lower per-element grad clip default from 5.0 to 1.0 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5440
* studio/frontend: drop unused next dependency by @danielhanchen in https://github.com/unslothai/unsloth/pull/5438
* Update version-compat-ci.yml by @rolandtannous in https://github.com/unslothai/unsloth/pull/5445
* ci: merge duplicate `with:` keys in notebooks-ci checkout steps by @rolandtannous in https://github.com/unslothai/unsloth/pull/5447
* studio/chat: built-in web search for OpenAI, Anthropic, OpenRouter, Kimi by @rolandtannous in https://github.com/unslothai/unsloth/pull/5443
* ci: make compiler-cache shim test order-independent by @danielhanchen in https://github.com/unslothai/unsloth/pull/5449
* Studio: o3 reasoning summary payload by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5426
* ci: compiler-cache-shim must mutate live module globals + skip rerun by @danielhanchen in https://github.com/unslothai/unsloth/pull/5452
* Polish/cloud to providers by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5450
* ci: cap each compiler-sweep iteration with SIGALRM + log progress by @danielhanchen in https://github.com/unslothai/unsloth/pull/5456
* ci: add tx >=5,<6 slow compile model_types to KNOWN_BROKEN_COMPILE by @danielhanchen in https://github.com/unslothai/unsloth/pull/5458
* Restore Flash > SDPA > Flex priority for non-gemma3 models by @mmathew23 in https://github.com/unslothai/unsloth/pull/5455
* ci: stop a partial mmproj cache from poisoning Mac Studio GGUF CI by @danielhanchen in https://github.com/unslothai/unsloth/pull/5459
* ci: make Windows Stop Studio teardown tolerate Git Bash signal exit by @danielhanchen in https://github.com/unslothai/unsloth/pull/5460
* Studio: make API key optional for local providers (llama.cpp/vLLM/Ollama) by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5457
* studio/chat: built-in code execution for OpenAI + Anthropic by @rolandtannous in https://github.com/unslothai/unsloth/pull/5461
* ci: switch Windows Stop Studio to a cmd no-op marker by @danielhanchen in https://github.com/unslothai/unsloth/pull/5462
* tests: raise pwsh/bash subprocess timeout from 10s to 60s by @danielhanchen in https://github.com/unslothai/unsloth/pull/5463
* studio/install: repair upstream llama.cpp prebuilt mangled symlinks by @danielhanchen in https://github.com/unslothai/unsloth/pull/5465
* studio/chat: OpenAI container picker delete reliability by @rolandtannous in https://github.com/unslothai/unsloth/pull/5466
* studio/install: strip top-level dir from repaired symlink target by @danielhanchen in https://github.com/unslothai/unsloth/pull/5467
* Stop: drop Ollama API key, clean up code execution UI by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5464
* tests/openai: patch httpx.AsyncClient ctor so delete tests hit mock by @danielhanchen in https://github.com/unslothai/unsloth/pull/5469
* revert: stop touching DEVICE_TYPE == cuda branches for CPU CI by @danielhanchen in https://github.com/unslothai/unsloth/pull/5473
* ci: drop `cache: 'npm'` from setup-node (silent abort on Windows) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5474
* ci: bump Mac json-images timeout 30 -> 45 min (cache-miss path) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5475
* ci: wrap hf download in xet-tuned stall-retry loop (root-cause Mac 30-min hang) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5476
* ci: deterministic check for studio/frontend dep removals by @danielhanchen in https://github.com/unslothai/unsloth/pull/5478
* studio/frontend: drop unused dependencies, move type pkg to devDeps by @danielhanchen in https://github.com/unslothai/unsloth/pull/5477
* intel-gpu: refresh xpu extras (fix torch 2.10, add 2.7.1 / 2.9.1 / 2.11.0 / 2.12.0) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5484
* Studio: auto-load models when adding a cloud provider by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5472
* Studio: code execution config visual polish  by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5471
* disable_torchcodec_if_broken: also patch datasets and clean sys.modules by @danielhanchen in https://github.com/unslothai/unsloth/pull/5483
* tests: pinned-symbol canary for unsloth-zoo save_pretrained_merged guards (#5410) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5433
* intel-gpu: pin unsloth_zoo>=2026.5.2 (fixes #5494) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5499
* fix(sentence_transformer): resume PEFT checkpoints under sentence-transformers >= 5.4 by @Etherll in https://github.com/unslothai/unsloth/pull/5454
* Studio: serialise GGUF reload and inherit unsloth-run extra args by @danielhanchen in https://github.com/unslothai/unsloth/pull/5427
* Studio: IME / multilingual composer regression test + RTL dir="auto" by @danielhanchen in https://github.com/unslothai/unsloth/pull/5485
* fix: preserve tokenizer eos token on merged saves by @anmolxlight in https://github.com/unslothai/unsloth/pull/5451
* studio/chat: reuse Anthropic code_execution container across turns by @rolandtannous in https://github.com/unslothai/unsloth/pull/5519
* Fix Linux prebuilt installs for branch-based llama.cpp releases by @mmathew23 in https://github.com/unslothai/unsloth/pull/5493
* Studio: stop hint, Uvicorn log rename, reachability check + Mac UI CI retry hardening by @danielhanchen in https://github.com/unslothai/unsloth/pull/5503
* Studio composer action pill styling by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5522
* Fix /recommended-folders 500 on unreadable model directories (Python 3.12+) by @mmathew23 in https://github.com/unslothai/unsloth/pull/5523
* studio/chat: persist Anthropic container id on first turn of new thread by @rolandtannous in https://github.com/unslothai/unsloth/pull/5526
* studio/openai: align chat completions docstring with stream=false default (closes #5047) by @wtfashwin in https://github.com/unslothai/unsloth/pull/5524
* Add a simple --version flag by @melroy89 in https://github.com/unslothai/unsloth/pull/5516
* studio: load cached GGUF models when fully offline by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5505
* studio: expose launcher capability bits on unauth /api/health by @danielhanchen in https://github.com/unslothai/unsloth/pull/5486
* studio: tighten sandbox blocklist precision (bash, hf upload, NOFILE) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5487
* studio: scope cancel-cleanup to in-flight tmp dirs; walk back tool_call_id by @danielhanchen in https://github.com/unslothai/unsloth/pull/5488
* studio: proxy-aware login rate-limit; allow google favicons in CSP by @danielhanchen in https://github.com/unslothai/unsloth/pull/5489
* studio/frontend: wire logout, singleflight refresh, shared 422 helper, current-password input by @danielhanchen in https://github.com/unslothai/unsloth/pull/5490
* tests/studio: lock in Windows GPU detection fix (#5106) with a synthetic CI test by @danielhanchen in https://github.com/unslothai/unsloth/pull/5376
* Studio: auto-enable MTP speculative decoding for MTP GGUFs by @danielhanchen in https://github.com/unslothai/unsloth/pull/5527
* Studio: warn when llama.cpp prebuilt is too old for MTP by @danielhanchen in https://github.com/unslothai/unsloth/pull/5528
* Studio: warn when llama.cpp prebuilt is at least 3 days behind by @danielhanchen in https://github.com/unslothai/unsloth/pull/5529
* studio: extend offline DNS auto-detect to inference parent + training by @danielhanchen in https://github.com/unslothai/unsloth/pull/5512
* Fix ORPO text-only tokenization with processors by @alkinun in https://github.com/unslothai/unsloth/pull/5501
* fix(studio/worker): inject --gcc-install-dir for HIP source builds on Ubuntu 24.04 by @h34v3nzc0dex in https://github.com/unslothai/unsloth/pull/5517
* studio: gate image input on effective vision capability by @Etherll in https://github.com/unslothai/unsloth/pull/5492
* studio/install: fix mac desktop shortcut spawning and lifecycle by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5496
* studio: add uninstall.sh and document it in README by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5497
* Studio update CI: round-trip install -> update -> uninstall by @danielhanchen in https://github.com/unslothai/unsloth/pull/5536
* studio: fix Connections dialog UX issues surfaced by image-gate probe by @danielhanchen in https://github.com/unslothai/unsloth/pull/5518
* studio: add uninstall.ps1 for Windows by @danielhanchen in https://github.com/unslothai/unsloth/pull/5513
* Fix num_logits_to_keep regression on transformers >= 4.52 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5538
* Uninstaller script by @PTFOPlayer in https://github.com/unslothai/unsloth/pull/4611
* Add OpenDocument chat attachments by @alkinun in https://github.com/unslothai/unsloth/pull/5510
* studio/frontend: stop showing Generating spinner on empty welcome view by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5530
* studio/frontend: swap Hugeicons spokes spinner for CSS ring by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5531
* studio/frontend: grow chat composer to 16 rows and inset scrollbar by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5540
* studio/frontend: make toast and inline error text selectable and copyable by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5506
* studio: add dismissable toasts with corner close button by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5509
* studio: install flash-linear-attention and tilelang for Qwen3.5 family by @danielhanchen in https://github.com/unslothai/unsloth/pull/5434
* studio/frontend: soften toast shadow and tighten vertical padding by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5511
* fast_generate: unify legacy/new logits kwarg + fix Mistral merge site by @danielhanchen in https://github.com/unslothai/unsloth/pull/5543
* studio/frontend: hide Current password input on first boot by @danielhanchen in https://github.com/unslothai/unsloth/pull/5545
* tests/studio: tighten MLX smoke gates (loss + round-trip, _on_step grad_norm) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5537
* tests + CI: callback signature drift detector by @danielhanchen in https://github.com/unslothai/unsloth/pull/5498
* images: use narrower Discord button and drop duplicate by @danielhanchen in https://github.com/unslothai/unsloth/pull/5552
* fix(studio): handle expired OpenAI shell-tool containers without surfacing error in chat by @rolandtannous in https://github.com/unslothai/unsloth/pull/5547
* studio/chat: release stuck IME flag when compositionend never fires by @wtfashwin in https://github.com/unslothai/unsloth/pull/5551

## New Contributors
* @Anai-Guo made their first contribution in https://github.com/unslothai/unsloth/pull/5329
* @uderbashi made their first contribution in https://github.com/unslothai/unsloth/pull/5168
* @Tenith01 made their first contribution in https://github.com/unslothai/unsloth/pull/5356
* @anmolxlight made their first contribution in https://github.com/unslothai/unsloth/pull/5451
* @wtfashwin made their first contribution in https://github.com/unslothai/unsloth/pull/5524
* @melroy89 made their first contribution in https://github.com/unslothai/unsloth/pull/5516
* @h34v3nzc0dex made their first contribution in https://github.com/unslothai/unsloth/pull/5517
* @PTFOPlayer made their first contribution in https://github.com/unslothai/unsloth/pull/4611

**Full Changelog**: https://github.com/unslothai/unsloth/compare/v0.1.39-beta...v0.1.40-beta

## What's Changed in Unsloth-Zoo
* Register Gemma-4 MoE LoRA extractor to fix grouped_mm contraction crash by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/624
* feat(mlx): Apple Silicon training (text + VLM, LoRA / full FT, CCE, export) by @Manan17 in https://github.com/unslothai/unsloth-zoo/pull/620
* tests: skip MoE LoRA extractor coverage when discovery finds zero classes by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/628
* tests: pivot MoE-coverage canary to _unsloth_already_patched marker by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/630
* fix(compiler): make higher_precision_softmax idempotent by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/631
* fix(mlx): unblock GGUF export and LoRA reload on Apple Silicon by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/627
* fix(compiler): unblock all model_types across transformers 4.57.6 and 5.x by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/632
* Mask for gemma3 attn by @Datta0 in https://github.com/unslothai/unsloth-zoo/pull/635
* Multi Image GRPO by @Datta0 in https://github.com/unslothai/unsloth-zoo/pull/613
* [GRPO] Try returning hidden statex for GRPO by @Datta0 in https://github.com/unslothai/unsloth-zoo/pull/609
* Refactor and consolidate moe lora extractors by @Datta0 in https://github.com/unslothai/unsloth-zoo/pull/629
* security + CI: mirror unsloth's hardening stack onto zoo (greenfield .github/) by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/637
* remove unsloth_zoo/import_fixes.py: redundant with unsloth's by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/639
* chore: trim verbose comments across PR #637 landing by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/640
* scripts: ship deterministic comment / docstring-only diff verifier by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/641
* fix mlx: Adds the MLX training path used by Studio on Apple Silicon by @mmathew23 in https://github.com/unslothai/unsloth-zoo/pull/634
* tests: drift detectors cover transformers 5.x (mirror unsloth PR #5423) by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/642
* gpt_oss: reorder helpers before patch_gpt_oss_bnb4bit_auto by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/643
* __init__: lazy-load legacy MLX aliases on every host by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/644
* tests: contain security-conftest network block; fix stale mlx paths; skip GPU import in trainer-exec-marker by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/648
* fix CI fallout from MLX subpackage refactor (#634) by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/646
* tests: tolerate transformers 5.x source/signature drift in two zoo drift detectors by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/650
* tests: skip _assert_params_superset when upstream forward is (*args, **kwargs) by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/651
* mlx: lower max_grad_value default from 5.0 to 1.0 by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/652
* saving: layout-aware MoE LoRA merge + loud-fail on fallback (#5410) by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/647
* tests: follow MoE merge wrapper delegation in drift detector by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/653
* additional import try except handling for mlx by @mmathew23 in https://github.com/unslothai/unsloth-zoo/pull/654
* Patch every LOSS_MAPPING key aliased to ForCausalLMLoss by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/656
* deps: bump torch upper cap to <2.13.0 (allow xpu 2.11.0 / 2.12.0) by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/658
* Auto-install fused lm_head + cross_entropy forward (opt-in) by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/657
* tests: CPU regression detectors for the MoE merge / save path (#5410) by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/655
* Fix GGUF converter sibling imports by @alkinun in https://github.com/unslothai/unsloth-zoo/pull/661
* fix embedding matrix size mismatch bug by @CodeMan62 in https://github.com/unslothai/unsloth-zoo/pull/645
* Honor UNSLOTH_RETURN_LOGITS in fused forward by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/665
* __init__: include HF_DATASETS_OFFLINE in the offline env cross-sync by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/664
* compiler: single-matmul opt-in for UNSLOTH_RETURN_LOGITS=1 by @danielhanchen in https://github.com/unslothai/unsloth-zoo/pull/666