v0.1.806-beta

Run Qwen3.8-Flash-Next and GLM-5.3-Flash up to 2x faster with MTP. MTP is enabled by default, you can still disable it.
Also our new release includes 170+ training, chat, hardware, and performance improvements.

## Highlights
*  **Smoother model loading** (less errors) across local servers and connected providers.
* Faster and less laggy UI with follow-up turns much faster for all chats.
*  **Safer chat edits** that preserve tool cards, reply details, and conversation branches.
* **New local media APIs** for video, audio, and MLX-served models.
* New audio support with new models, progress tracking including: MiniMax-Music3, Higgs, MOSS and more!
* Improved multi-GPU planning, memory fitting, and split-model training.
* Strengthened AMD/ROCm detection, installation, and GPU compatibility.
* Upgraded MCP, Deep Research, OAuth, and agent tool reliability.  

## Qwen3.8-Flash + GLM-5.3-Flash
- Qwen and GLM now generate faster with MTP enabled by default.
- Use GLM tools across longer, multi-turn chats.
- Qwen automatically applies the recommended settings for thinking and non-thinking modes.

Download [Qwen3.8-Flash-Next](https://huggingface.co/unsloth/Qwen3.8-Flash-Next) and [GLM-5.3-Flash](https://huggingface.co/unsloth/GLM-5.3-Flash). See the [Qwen guide](https://unsloth.ai/docs/models/qwen3.8-next) and [GLM guide](https://unsloth.ai/docs/models/glm-5.3-flash) for recommended settings and available GGUFs.

## Faster MLX inference

- Fine-tune both large MoE models with text or images on Apple Silicon using MLX.
- Long Qwen chats now run much faster on Mac, with follow-up turns up to 30x faster.
- MLX models now use their full context size and support much longer batched generation.
- MLX releases GPU memory more cleanly between generation bursts and model switches.
- Serve MLX models through Unsloth's OpenAI-compatible API.

## Audio

* Added support for **MiniMax-Music3, Higgs, MOSS audio models**. 
* Added **live progress updates** while audio is being generated. 
* Audio clips can now be **archived and managed**. 
* Improved reliability with **custom TTS playback fixes, Whisper pairing checks, and stronger audio testing**.   

## Chat + tools

- Run several tool calls at once without mixing up their arguments.
- Keep tools available when chatting with images.
- Each chat keeps its MCP connection for faster tool calls.
- Local models can edit code using Codex’s apply_patch tool.
- Continue long chats with images and other media using Auto Compaction.
- Review and approve Deep Research plans before research starts.

## Training + hardware

- Train larger models across multiple GPUs with automatic placement.
- AMD installs choose the best build across Windows and Linux, with BF16 on more GPUs.
- Export GLM-5.3 MLX fine-tunes to GGUF.
- Choose custom GGUF shard sizes and save locations.

## API + Desktop

- Generate videos through the new OpenAI-compatible Videos API.
- Updates download in the background and install when you restart.
- Choose a custom port for LAN access.
- Generate audio with Higgs, MOSS and MiniMax models.
- Track audio generation progress and archive finished clips.
- Model downloads show clearer progress and can switch from Xet to HTTP automatically.

## Download Unsloth Desktop

Unsloth Desktop is **free and open source**. Download it for:

- **Windows**
- **macOS**
- **Linux**

**[🦥 Download Unsloth Desktop](https://unsloth.ai/download)**

## What's Changed
* Put the smart offload planner back behind its flag by @danielhanchen in https://github.com/unslothai/unsloth/pull/9862
* Studio: stop the per-chunk autosave writing back messages the server owns by @danielhanchen in https://github.com/unslothai/unsloth/pull/9865
* Bump install.sh / install.ps1 pins to unsloth>=2026.8.22 by @danielhanchen in https://github.com/unslothai/unsloth/pull/9868
* Fix Studio hydrating synced GGUF files before selection by @milewski in https://github.com/unslothai/unsloth/pull/9539
* Fix datasets PyArrow registration after failed warm import by @Imagineer99 in https://github.com/unslothai/unsloth/pull/9626
* fix(studio): stop offering a transformers upgrade where it cannot load anything by @Lyxot in https://github.com/unslothai/unsloth/pull/9878
* Fix Studio CPT overwriting LFM2 all-linear LoRA targets by @Souravrajvi0 in https://github.com/unslothai/unsloth/pull/9870
* fix(studio): accept trailing slash for model discovery by @Imagineer99 in https://github.com/unslothai/unsloth/pull/9876
* studio: accept shared youtube clipboard links by @mahiatlinux in https://github.com/unslothai/unsloth/pull/9903
* studio: tell users when a canvas is blocked by the network access setting by @NilayYadav in https://github.com/unslothai/unsloth/pull/9894
* studio: fix streamed tool calls in studio api monitor by @Etherll in https://github.com/unslothai/unsloth/pull/9896
* studio: serve installed mlx models through the openai-compatible api by @mahiatlinux in https://github.com/unslothai/unsloth/pull/8768
* Fix sidebar footer width and context menu radius by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/9918
* strip inherited claude provider routing from local agent launches by @Souravrajvi0 in https://github.com/unslothai/unsloth/pull/9871
* studio: honor a cancelled load inside the llama-server health wait by @NilayYadav in https://github.com/unslothai/unsloth/pull/9893
* fix model hub recent sorting for on-device models by @Imagineer99 in https://github.com/unslothai/unsloth/pull/9642
* add a setting that tells the model the current date by @mahiatlinux in https://github.com/unslothai/unsloth/pull/8879
* studio: fix lan address actions and launch bind reporting by @Lyxot in https://github.com/unslothai/unsloth/pull/9170
* rag: avoid re-hashing linked-folder snapshots in start_ingestion by @DhruvaMyakeri in https://github.com/unslothai/unsloth/pull/9913
* fix(studio): coalesce concurrent keyless-settings reads to stop SQLite thundering herd by @Greninja44 in https://github.com/unslothai/unsloth/pull/9911
* Studio: add the OpenAI videos API on /v1/videos by @NilayYadav in https://github.com/unslothai/unsloth/pull/9891
* Clarify tool approval copy by @Imagineer99 in https://github.com/unslothai/unsloth/pull/8738
* Pin Claude Code to the loaded model's real context window by @NilayYadav in https://github.com/unslothai/unsloth/pull/9940
* studio: drop the schema bounds llama.cpp's grammar engine cannot compile by @Lyxot in https://github.com/unslothai/unsloth/pull/9923
* Fix shell selection for remote Agents commands by @Imagineer99 in https://github.com/unslothai/unsloth/pull/8844
* Studio: accept the loadConfig keys a saved system-prompt preset already sends by @indrajeetapache in https://github.com/unslothai/unsloth/pull/9881
* Studio: compact long chats with media attachments by @Imagineer99 in https://github.com/unslothai/unsloth/pull/9476
* Studio: add Higgs, MOSS, and MiniMax audio models by @Etherll in https://github.com/unslothai/unsloth/pull/8794
* studio: add ordered arguments to mcp server settings by @Etherll in https://github.com/unslothai/unsloth/pull/9943
* studio: list route-aware models through the openai api by @NilayYadav in https://github.com/unslothai/unsloth/pull/9892
* fix(studio): repair custom TTS endpoint test and playback by @Souravrajvi0 in https://github.com/unslothai/unsloth/pull/9905
* Studio: render escaped inline math in lists by @Imagineer99 in https://github.com/unslothai/unsloth/pull/9796
* deprecate unsupported custom prompt templates by @vineethsaivs in https://github.com/unslothai/unsloth/pull/8938
* Restore UNSLOTH_RETURN_LOGITS after prediction_step instead of forcing it to 0 by @danielhanchen in https://github.com/unslothai/unsloth/pull/6239
* Studio: update in the background and restart when ready by @NilayYadav in https://github.com/unslothai/unsloth/pull/9890
* Serve every MiniMax H3 asset from unsloth rather than a community repack by @danielhanchen in https://github.com/unslothai/unsloth/pull/8832
* fix(studio): gate external tool nudges and preserve retry context by @Biotrioo in https://github.com/unslothai/unsloth/pull/9125
* Chat: let the LoRA compare rehydrate a pair the generalized compare saved by @lonexreb in https://github.com/unslothai/unsloth/pull/9908
* fix: expose codex apply_patch for local models (#9114) by @Biotrioo in https://github.com/unslothai/unsloth/pull/9121
* Fix duplicate GGUF rows in custom model folders by @Imagineer99 in https://github.com/unslothai/unsloth/pull/9975
* Revert "fix(studio): gate external tool nudges and preserve retry context" by @danielhanchen in https://github.com/unslothai/unsloth/pull/9990
* studio: isolate mcp browser test state by @mahiatlinux in https://github.com/unslothai/unsloth/pull/9951
* Fix three CI regressions: code-span anchor, load-dispatch anchor, recordless pip install by @danielhanchen in https://github.com/unslothai/unsloth/pull/9959
* studio: fix audio ci tests by @mahiatlinux in https://github.com/unslothai/unsloth/pull/9953
* studio: fix openai route ci tests by @mahiatlinux in https://github.com/unslothai/unsloth/pull/9952
* Skip the inline-code mask when every backtick is inside a fence by @danielhanchen in https://github.com/unslothai/unsloth/pull/9991
* stop the planner withholding a split reserve from one card, and price physical cores by @danielhanchen in https://github.com/unslothai/unsloth/pull/9863
* Parse the downloaded model mapper instead of exec'ing it by @danielhanchen in https://github.com/unslothai/unsloth/pull/9777
* Recapture the control leg reference against the fixed loss normalisation by @danielhanchen in https://github.com/unslothai/unsloth/pull/9569
* Stop os.dup2 alone being a reverse shell, and allowlist the zoo URL guard by @danielhanchen in https://github.com/unslothai/unsloth/pull/9994
* Give the in-venv CLI tests a writable STUDIO_HOME by @danielhanchen in https://github.com/unslothai/unsloth/pull/9993
* Settle the compact layout before measuring the permission pill by @danielhanchen in https://github.com/unslothai/unsloth/pull/9992
* Fail CI on a new exec/eval/compile of a value that is not written out by @danielhanchen in https://github.com/unslothai/unsloth/pull/9999
* Advertise a backend version a released AppImage will accept by @danielhanchen in https://github.com/unslothai/unsloth/pull/10001
* Budget the dyld classifier in CPU time, not wall clock by @danielhanchen in https://github.com/unslothai/unsloth/pull/10003
* Stop one slow SQLite writer from becoming a stream of Studio log errors by @danielhanchen in https://github.com/unslothai/unsloth/pull/9996
* studio: cache the servability scan behind GET /v1/models by @danielhanchen in https://github.com/unslothai/unsloth/pull/9998
* Make the multi-turn smoke actually assert that history reached the model by @danielhanchen in https://github.com/unslothai/unsloth/pull/10009
* Share the fixture backend version with the model-download E2E by @danielhanchen in https://github.com/unslothai/unsloth/pull/10002
* Give the cancel drain a budget sized for a loaded runner by @danielhanchen in https://github.com/unslothai/unsloth/pull/10008
* Size the model picker's row metadata to the ink it holds by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/10006
* Studio: fix unscrollable model dropdowns on the export page by @Lyxot in https://github.com/unslothai/unsloth/pull/10020
* Ground the multi-turn smoke on the turn that survives a 270M model by @danielhanchen in https://github.com/unslothai/unsloth/pull/10025
* Studio: honor the effort level Claude Code sends on /v1/messages by @NilayYadav in https://github.com/unslothai/unsloth/pull/9941
* Studio: archive audio clips by @NilayYadav in https://github.com/unslothai/unsloth/pull/9637
* Pre-flight whisper pairing before the chained install by @danielhanchen in https://github.com/unslothai/unsloth/pull/10027
* Dim an over budget model row instead of pilling every line by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/10007
* studio: reap durable chat generation runs that stop making progress by @danielhanchen in https://github.com/unslothai/unsloth/pull/9997
* Stop three Backend CI tests measuring the runner instead of the code by @danielhanchen in https://github.com/unslothai/unsloth/pull/10035
* Let the AppImage E2E use the Retry the failed quant row offers by @danielhanchen in https://github.com/unslothai/unsloth/pull/10032
* Keep a modules_to_save copy on the card it was loaded onto by @danielhanchen in https://github.com/unslothai/unsloth/pull/10034
* Train a model the planner split across cards, instead of dying in index_select by @danielhanchen in https://github.com/unslothai/unsloth/pull/9995
* Name the encoding when the offload test reads llama.py by @danielhanchen in https://github.com/unslothai/unsloth/pull/10039
* Answer llama-server discovery probes instead of serving them the app shell by @danielhanchen in https://github.com/unslothai/unsloth/pull/10037
* Ask one formula whether a GGUF fits by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/10029
* Stop the llama-server probe deny-list answering 405 by @danielhanchen in https://github.com/unslothai/unsloth/pull/10049
* Stop the subagent tool pre-approval swallowing the forwarded prompt by @NilayYadav in https://github.com/unslothai/unsloth/pull/9939
* Use the plain download icon in the Hub On Device views by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/10055
* Tag the compat endpoint's Ollama rows with their real source by @lonexreb in https://github.com/unslothai/unsloth/pull/9988
* Stop the GGUF header and the chat detail polls from filling the log by @danielhanchen in https://github.com/unslothai/unsloth/pull/10040
* feat(export): GGUF shard size control, save directory override, fix private Hub push by @LeoBorcherding in https://github.com/unslothai/unsloth/pull/6107
* Studio: set `GGML_CUDA_ENABLE_UNIFIED_MEMORY` only where it gains memory by @oobabooga in https://github.com/unslothai/unsloth/pull/9884
* Wait for the idle batch flush instead of racing a fixed sleep by @danielhanchen in https://github.com/unslothai/unsloth/pull/10063
* Use the head-aware planner for Studio's multi-GPU device map by @danielhanchen in https://github.com/unslothai/unsloth/pull/10042
* Let tool chats decode in parallel instead of one at a time by @danielhanchen in https://github.com/unslothai/unsloth/pull/10046
* Fix excessive studio.db writes during chat inference by @Imagineer99 in https://github.com/unslothai/unsloth/pull/9945
* Stop shipping test suites in the wheel and the sdist by @danielhanchen in https://github.com/unslothai/unsloth/pull/10060
* Studio: keep a Windows update from leaving a CPU-only PyTorch by @danielhanchen in https://github.com/unslothai/unsloth/pull/9857
* Read the chat storage keys on call, not at module scope by @danielhanchen in https://github.com/unslothai/unsloth/pull/9844
* Studio: install torch from an index that has kernels for the user's AMD GPU by @oobabooga in https://github.com/unslothai/unsloth/pull/9829
* fix(studio): normalize Muse Glimmer's recipient-addressed protocol by @Lyxot in https://github.com/unslothai/unsloth/pull/8422
* fix(studio): read tool-call arguments as the types their schema declares by @Lyxot in https://github.com/unslothai/unsloth/pull/10019
* Studio: notice a quarantined payload before trusting the update fast path by @danielhanchen in https://github.com/unslothai/unsloth/pull/10053
* Say why Studio could not enumerate a GPU, instead of just logging [] by @danielhanchen in https://github.com/unslothai/unsloth/pull/10056
* CI: fix llama.cpp seed caching, Diffusers wheel installs, and isolated test coverage by @oobabooga in https://github.com/unslothai/unsloth/pull/9979
* Studio: resume HTTP after Xet stall without a transport-conflict banner by @Souravrajvi0 in https://github.com/unslothai/unsloth/pull/10045
* Fix GLM-5.3 chats breaking after a tool runs by @NilayYadav in https://github.com/unslothai/unsloth/pull/9938
* Studio: fix clipped settings scrollbar on Windows WebView2 by @NilayYadav in https://github.com/unslothai/unsloth/pull/10074
* Bump dompurify from 3.4.2 to 3.4.14 in /studio/frontend in the npm-frontend-security group across 1 directory by @dependabot[bot] in https://github.com/unslothai/unsloth/pull/6753
* Studio: show progress during audio generation by @Etherll in https://github.com/unslothai/unsloth/pull/10081
* Show activity on a 0-byte Xet download and keep finished cards visible longer by @NilayYadav in https://github.com/unslothai/unsloth/pull/9303
* Fix Hub On Device Selection Across Refresh and Download Changes by @Sneakr in https://github.com/unslothai/unsloth/pull/9922
* Patch Accelerator.distributed_type as a property, not a bare function by @vineethsaivs in https://github.com/unslothai/unsloth/pull/10028
* Stop charging a chat the whole cache because Max Tokens says "Max" by @danielhanchen in https://github.com/unslothai/unsloth/pull/10070
* Stop two installer tests answering about the runner instead of the code by @danielhanchen in https://github.com/unslothai/unsloth/pull/10079
* Studio: split parallel tool-call arguments at JSON object boundaries by @danielhanchen in https://github.com/unslothai/unsloth/pull/10059
* Move the studiobench attribution config in with the harness by @danielhanchen in https://github.com/unslothai/unsloth/pull/10100
* Fix Accelerator.distributed_type bound-method bug on single-GPU patch by @mayuriphad in https://github.com/unslothai/unsloth/pull/10072
* Wrap Accelerator.distributed_type patch in property() (Closes #10016) by @mihiii1337 in https://github.com/unslothai/unsloth/pull/10087
* Add configurable LAN access port by @wasimysaid in https://github.com/unslothai/unsloth/pull/10082
* Add GGUF fit-target control and wire to llama-server --fit-target by @aiSynergy37 in https://github.com/unslothai/unsloth/pull/4882
* Use the public fastmcp call_tool so the MCP guard test survives fastmcp 4 by @danielhanchen in https://github.com/unslothai/unsloth/pull/10110
* Scope the llama.cpp update test's Popen patch to the installer by @danielhanchen in https://github.com/unslothai/unsloth/pull/10105
* Run the llama.cpp keep-install back-compat tests on Windows by @danielhanchen in https://github.com/unslothai/unsloth/pull/10106
* Stop the torchao ROCm test from depending on the host's GPU by @danielhanchen in https://github.com/unslothai/unsloth/pull/10099
* fix(studio): show the confirmation card the argument text the tool is run with by @Lyxot in https://github.com/unslothai/unsloth/pull/10023
* Reduce the comments added by PR #9979 by @danielhanchen in https://github.com/unslothai/unsloth/pull/10107
* Give the unsloth_zoo pytest step a per-test timeout by @danielhanchen in https://github.com/unslothai/unsloth/pull/10108
* Studio: report a context size for models loaded through MLX by @NilayYadav in https://github.com/unslothai/unsloth/pull/10098
* Studio: tell the browser about the bootstrap shutdown deadline by @danielhanchen in https://github.com/unslothai/unsloth/pull/10104
* AMD: Fix BF16 detection on gfx10 ROCm GPUs by @sunjingwen21 in https://github.com/unslothai/unsloth/pull/7682
* fix(amd): gate training GPU selection on the installed torch build's arch coverage by @LeoBorcherding in https://github.com/unslothai/unsloth/pull/8791
* Linux: fix ROCm detection on split Debian stacks by @TaskPuppyNatani in https://github.com/unslothai/unsloth/pull/8886
* Studio: send tool results to the model in the right order by @NilayYadav in https://github.com/unslothai/unsloth/pull/10091
* studio: keep MTP on for GLM-5.3-Flash by @danielhanchen in https://github.com/unslothai/unsloth/pull/10119
* Studio: report the GPUs the OS sees, not just the ones PyTorch opened by @danielhanchen in https://github.com/unslothai/unsloth/pull/9858
* Apply the context length and quant settings when attaching to a running server by @NilayYadav in https://github.com/unslothai/unsloth/pull/10093
* Reuse the MLX prompt cache on hybrid recurrent models by @Lyxot in https://github.com/unslothai/unsloth/pull/10062
* Report a trainer kwarg that neither the trainer nor the config takes by @vineethsaivs in https://github.com/unslothai/unsloth/pull/9944
* Studio: keep a Downloads entry when the list is empty by @Souravrajvi0 in https://github.com/unslothai/unsloth/pull/9849
* Kaggle T4: Latest_compile and Vision_FLA_compile legs, plus gpt-oss completions and GGUF by @danielhanchen in https://github.com/unslothai/unsloth/pull/9700
* Kaggle CI: use both accounts, weighted by their own weekly quota by @danielhanchen in https://github.com/unslothai/unsloth/pull/10121
* Kaggle T4: split the load phase into fetch and weight load by @danielhanchen in https://github.com/unslothai/unsloth/pull/9692
* fix(amd): don't enable torch.compile when Triton's C toolchain can't compile it (Windows) by @LeoBorcherding in https://github.com/unslothai/unsloth/pull/7704
* fix(AMD): route Fedora/Bazzite to per-arch ROCm wheels when no version is readable by @LeoBorcherding in https://github.com/unslothai/unsloth/pull/9152
* Studio: use plain arrow icons for up, down, left and right by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/10129
* Prune an open PR's caches once nothing has read them by @danielhanchen in https://github.com/unslothai/unsloth/pull/10123
* Cache the same Playwright engines under the same key by @danielhanchen in https://github.com/unslothai/unsloth/pull/10122
* Make the notebooks smoke job reach the check it exists for by @danielhanchen in https://github.com/unslothai/unsloth/pull/10125
* Remove OpenSSF Scorecard workflow by @danielhanchen in https://github.com/unslothai/unsloth/pull/10133
* feat(studio): give MLX models a real context length, from the model config to the KV cache by @Lyxot in https://github.com/unslothai/unsloth/pull/8125
* Bump the actions group with 3 updates by @dependabot[bot] in https://github.com/unslothai/unsloth/pull/9797
* Bump the cargo-tauri group across 1 directory with 6 updates by @dependabot[bot] in https://github.com/unslothai/unsloth/pull/9798
* Stop `unsloth start claude` from opening with a model it cannot use by @NilayYadav in https://github.com/unslothai/unsloth/pull/10089
* studio: fall back to the MTP/ folder when a repo publishes no root drafter by @danielhanchen in https://github.com/unslothai/unsloth/pull/10124
* Studio: auto-generate an admin password for headless public launches by @danielhanchen in https://github.com/unslothai/unsloth/pull/7392
* Studio: keep one MCP connection per chat instead of reconnecting every call by @NilayYadav in https://github.com/unslothai/unsloth/pull/10095
* Studio: restore Deep Research plan approval by @wasimysaid in https://github.com/unslothai/unsloth/pull/10134
* Studio: line up On Device rows, and list partial downloads by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/10128
* Widen model picker panel and header controls by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/10132
* Studio: keep download restart toasts job-scoped by @wasimysaid in https://github.com/unslothai/unsloth/pull/10136
* MLX: drain GPU work before clearing the cache by @Lyxot in https://github.com/unslothai/unsloth/pull/9921
* Fix stale Qwen sampling defaults after Studio upgrades by @Imagineer99 in https://github.com/unslothai/unsloth/pull/9643
* fix(studio): enforce Hub token caller boundary by @Imagineer99 in https://github.com/unslothai/unsloth/pull/10076
* Studio: keep tools available when a picture is attached by @NilayYadav in https://github.com/unslothai/unsloth/pull/10092
* Fix pre-commit formatting drift after #10076 by @Imagineer99 in https://github.com/unslothai/unsloth/pull/10154
* desktop: check for updates periodically by @mahiatlinux in https://github.com/unslothai/unsloth/pull/9505
* studio: use generation usage for conversation search budgets by @mahiatlinux in https://github.com/unslothai/unsloth/pull/9438
* studio: fetch the XPU triton wheel with --dest, not pip's -d by @Sletch in https://github.com/unslothai/unsloth/pull/10073
* fix(studio): poll inference status while CLI model is loading by @jimdawdy-hub in https://github.com/unslothai/unsloth/pull/5901
* Studio: restore sticky checkpoint boundaries on tool-heavy threads by @alkinun in https://github.com/unslothai/unsloth/pull/9971
* Studio: surface every live provider model in the picker by @danielhanchen in https://github.com/unslothai/unsloth/pull/5684
* Fix llama.cpp Studio update failures on GitHub API rate limits (#9970) by @Souravrajvi0 in https://github.com/unslothai/unsloth/pull/9980
* Studio: dispatch on OpenAI agentic web_search action variants by @danielhanchen in https://github.com/unslothai/unsloth/pull/5787
* studio: tighten MTP reload guards and asymmetric spec flags for #5582 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5696
* Studio: don't re-prompt after model produced a complete answer by @danielhanchen in https://github.com/unslothai/unsloth/pull/5714
* Studio: don't size attached documents by a local model on hosted chats by @NilayYadav in https://github.com/unslothai/unsloth/pull/10164
* Studio: keep tool cards in place when you edit a reply by @NilayYadav in https://github.com/unslothai/unsloth/pull/10162
* Studio: keep a reply's details when you edit its text by @NilayYadav in https://github.com/unslothai/unsloth/pull/10161
* Studio: keep both branches when a chat's first message is edited by @NilayYadav in https://github.com/unslothai/unsloth/pull/10163
* Stop the kwarg-spacing post-pass dying on Python 3.11 by @danielhanchen in https://github.com/unslothai/unsloth/pull/10137
* Studio: stop Deep Research from throwing away a finished report by @NilayYadav in https://github.com/unslothai/unsloth/pull/10166
* Studio: show what changed in llama.cpp updates by @oobabooga in https://github.com/unslothai/unsloth/pull/10117
* Studio: test that a chat message keeps its identity (#9984) by @umran666 in https://github.com/unslothai/unsloth/pull/10064
* Reflow comments in studio/backend/core/inference to 120 columns by @danielhanchen in https://github.com/unslothai/unsloth/pull/10115
* Trim comments in studio/backend core services and the unsloth package by @danielhanchen in https://github.com/unslothai/unsloth/pull/10116
* Trim comments in the studiobench harness and unsloth_cli by @danielhanchen in https://github.com/unslothai/unsloth/pull/10118
* Studio: fix OAuth sign-in for MCP servers like Notion by @NilayYadav in https://github.com/unslothai/unsloth/pull/10167
* Studio: align Run Settings and titlebar controls by @alkinun in https://github.com/unslothai/unsloth/pull/10175
* Trim comments in studio/backend routes, utils and installers by @danielhanchen in https://github.com/unslothai/unsloth/pull/10112
* Update CODEOWNERS for current maintainers and code areas by @danielhanchen in https://github.com/unslothai/unsloth/pull/10191
* Security audit: re-approve fastmcp-slim 4.0.x and unsloth-zoo 2026.8.17 by @danielhanchen in https://github.com/unslothai/unsloth/pull/10187
* Stop the research cancellation tests failing on a busy CI runner by @danielhanchen in https://github.com/unslothai/unsloth/pull/10190
* Fix the frontend unit tests broken by a merge skew between #10161 and #10162 by @danielhanchen in https://github.com/unslothai/unsloth/pull/10186
* Fix the two stale source-text contract tests blocking every PR by @danielhanchen in https://github.com/unslothai/unsloth/pull/10185
* Fix Studio preferring MTP sidecars over embedded heads by @wasimysaid in https://github.com/unslothai/unsloth/pull/10181

## New Contributors
* @milewski made their first contribution in https://github.com/unslothai/unsloth/pull/9539
* @DhruvaMyakeri made their first contribution in https://github.com/unslothai/unsloth/pull/9913
* @Greninja44 made their first contribution in https://github.com/unslothai/unsloth/pull/9911
* @mayuriphad made their first contribution in https://github.com/unslothai/unsloth/pull/10072
* @mihiii1337 made their first contribution in https://github.com/unslothai/unsloth/pull/10087
* @aiSynergy37 made their first contribution in https://github.com/unslothai/unsloth/pull/4882
* @sunjingwen21 made their first contribution in https://github.com/unslothai/unsloth/pull/7682
* @TaskPuppyNatani made their first contribution in https://github.com/unslothai/unsloth/pull/8886
* @Sletch made their first contribution in https://github.com/unslothai/unsloth/pull/10073

**Full Changelog**: https://github.com/unslothai/unsloth/compare/v0.1.804-beta...v0.1.806-beta