v0.1.905-beta

We're introducing Windows, Mac and Linux sandboxing in Unsloth! [Read our guide](https://unsloth.ai/docs/new/studio/sandboxing-in-unsloth). Turn any text or vision LLM into a Jev-style decision model in Unsloth, with decision accuracy going from 30% to 80%. Train, test, export and serve decision models directly from Unsloth. Also included: native ComfyUI models, diffusion improvements and a better Browser in Desktop.

## Highlights

- **8th Oct Fixes** - Better Browser + 100+ bug fixes + perf fixes
- **Sandboxing** with Bwrap for Linux, Seatbelt for Mac and MXC for Windows
- **Turn any model into a Jev-style decision model.** Accuracy went from 30% to 80%
- **Load ComfyUI diffusion models** natively in Unsloth
- **Faster + more accurate diffusion** with INT8 ConvRot and more

https://github.com/user-attachments/assets/92959172-d2a2-450b-a5bf-6369a1e5ae16

## Decision models

- Train any text or vision LLM as a Jev-style decision model using QLoRA.
- Test trained models directly from the Decision API settings.
- Make decisions with confidence scores for every option.
- Export Clef models with Qwen3.5 backbones and Laya models to GGUF.
- Serve supported decision models through llama.cpp, including models that understand images.
- Save and resume smaller adapter and decision-head checkpoints.
- Guide at https://unsloth.ai/docs/basics/train-your-own-decision-model-with-unsloth
<img width="600" height="773" alt="image" src="https://github.com/user-attachments/assets/963eab71-49d1-47e7-9ccd-f3c5f1fc8943" />

## Diffusion + ComfyUI

- Run supported ComfyUI image and video models directly from Hugging Face.
- Unsloth now recognises ComfyUI checkpoints and local model folders automatically.
- Use your ComfyUI text encoders and VAEs in Unsloth.
- Run Krea-2, HunyuanImage-2.1 and Wan2.2 expert pairs, plus ComfyUI NVFP4 and MXFP8 models.
- Qwen-Image-2.1 now keeps more full-precision image detail with INT8 ConvRot enabled by default.
- Faster Qwen-Image-2.1 ConvRot generation on supported NVIDIA GPUs.

## Training + performance

- Train Qwen3.5-35B-A3B up to 4.1x faster and Qwen3-30B-A3B up to 3.3x faster with QLoRA on A100 and RTX PRO 6000.
- Improved sample packing for gated-delta, Mamba2 and short-convolution models.
- Train prompt and completion message lists as one conversation.
- Vision datasets now keep each row's own question.
- Chat exports and training data now include the system prompt.

## Browser + Desktop

- Ask about open pages in the Desktop Browser.
- Confirm Browser downloads and choose where files are saved.
- Reorder pinned pages in the sidebar like chats.
- Search continues past unusable results and can fall back to Wikipedia.
- Reply citations such as `[1]` now open as links.
- Pick, pin or unload RAG embedding models directly from the RAG menu.

## Sandboxing

- Bwrap on Linux, Seatbelt on Mac and MXC on Windows sandbox code the model runs.
- View sandbox status and choose protection levels in Settings.

## Download Unsloth Desktop

Unsloth Desktop is **free and open source**. Download it for:
<table>
  <tr>
    <td><b>Platform</b></td>
    <td><b>Link</b></td>
  </tr>
  <tr>
    <td><b>Windows</b></td>
    <td><a href='https://github.com/unslothai/unsloth/releases/latest/download/Unsloth-Desktop-Windows.exe'>Download</a></td>
  </tr>
  <tr>
    <td><b>macOS</b></td>
    <td><a href='https://github.com/unslothai/unsloth/releases/latest/download/Unsloth-Desktop-MacOS.dmg'>Download</a></td>
  </tr>
  <tr>
    <td><b>Linux x64 / Ubuntu (deb)</b></td>
    <td><a href='https://github.com/unslothai/unsloth/releases/latest/download/Unsloth-Desktop-Ubuntu.deb'>Download</a></td>
  </tr>
  <tr>
    <td><b>Linux ARM64 / Ubuntu 24.04+ (deb)</b></td>
    <td><a href='https://github.com/unslothai/unsloth/releases/latest/download/Unsloth-Desktop-Ubuntu-ARM64.deb'>Download</a></td>
  </tr>
  <tr>
    <td><b>Linux x64 (AppImage)</b></td>
    <td><a href='https://github.com/unslothai/unsloth/releases/latest/download/Unsloth-Desktop-Linux.AppImage'>Download</a></td>
  </tr>
  <tr>
    <td><b>Windows ARM64</b></td>
    <td><a href='https://github.com/unslothai/unsloth/releases/latest/Unsloth-Desktop-Windows-ARM64.exe'>Download</a></td>
  </tr>
</table>

## What's Changed
* Bump install.sh / install.ps1 pins to unsloth>=2026.10.1, unsloth-zoo>=2026.10.1 by @danielhanchen in https://github.com/unslothai/unsloth/pull/12869
* Studio: list embeddinggemma-2 first in the embedding model picker by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/12870
* Repair three checks that went red on main with the 10-06 Studio merges by @danielhanchen in https://github.com/unslothai/unsloth/pull/12868
* Studio: run ComfyUI-format video quants on the int8 / fp8 runtimes by @danielhanchen in https://github.com/unslothai/unsloth/pull/12851
* Studio: free PyAV's per-thread scalers before a fork so preexec_fn spawns still exec by @danielhanchen in https://github.com/unslothai/unsloth/pull/12863
* Tests: give the setup.ps1 download progress pwsh its own startup cache by @danielhanchen in https://github.com/unslothai/unsloth/pull/12882
* Studio: fill the Hebrew and Swedish strings that left the strict i18n check red on main by @danielhanchen in https://github.com/unslothai/unsloth/pull/12881
* Baseline the eight unsloth-zoo 2026.10.1 findings after review by @danielhanchen in https://github.com/unslothai/unsloth/pull/12884
* Support every PEFT init_lora_weights option, with fast PiSSA and MiCA init by @Suchitra-idu in https://github.com/unslothai/unsloth/pull/6879
* Sandbox test: wait for the cache scan workers before counting them by @danielhanchen in https://github.com/unslothai/unsloth/pull/12896
* Studio: keep browser panel tooltips, toasts and menus visible over desktop web pages by @oobabooga in https://github.com/unslothai/unsloth/pull/12895
* Studio: continue searching past unusable results and add a Wikipedia fallback by @oobabooga in https://github.com/unslothai/unsloth/pull/12892
* Studio: list every Transcribe ASR model in Voice settings by @Etherll in https://github.com/unslothai/unsloth/pull/12898
* Frontend test: give the cold Vite SSR render in reasoning-source-render room on a loaded runner by @danielhanchen in https://github.com/unslothai/unsloth/pull/12903
* Run shell suites and Windows browser checks in parallel by @oobabooga in https://github.com/unslothai/unsloth/pull/12899
* Studio: add a New badge beside Audio in the sidebar by @Etherll in https://github.com/unslothai/unsloth/pull/12891
* Composer settings driver: poll the submitted list instead of reading it once after the key press by @danielhanchen in https://github.com/unslothai/unsloth/pull/12920
* Studio: support current native builds and preserve CPU asset selection by @oobabooga in https://github.com/unslothai/unsloth/pull/12902
* Studio: pick the quant of a GGUF dictation model in Voice settings by @Etherll in https://github.com/unslothai/unsloth/pull/12900
* fix(studio): stop a managed runtime when the client drops its stream by @goodmai in https://github.com/unslothai/unsloth/pull/12266
* Studio: train prompt/completion message lists as one conversation by @NilayYadav in https://github.com/unslothai/unsloth/pull/12910
* Studio: keep each row's own question when training on a vision dataset by @NilayYadav in https://github.com/unslothai/unsloth/pull/12909
* Studio: train transparent PNG and WebP images on white instead of black by @NilayYadav in https://github.com/unslothai/unsloth/pull/12908
* Use the requested max_seq_length for encoder embedding models by @NilayYadav in https://github.com/unslothai/unsloth/pull/12915
* Studio: show the LAN address on the API page when LAN access is on by @NilayYadav in https://github.com/unslothai/unsloth/pull/12906
* Studio: hide the negative prompt on image models that ignore it by @NilayYadav in https://github.com/unslothai/unsloth/pull/12914
* Keep the notebook and saved model when unsloth-run runs a URL by @NilayYadav in https://github.com/unslothai/unsloth/pull/12907
* Studio: drag pinned pages in the sidebar like chats by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/12927
* Studio: Ask about this page on every page, desktop included by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/12926
* Docker: allow unsloth_root_shim.py into the build context by @danielhanchen in https://github.com/unslothai/unsloth/pull/12929
* Tauri transport test: start the late backend after the old ladder is spent, not at 3s by @danielhanchen in https://github.com/unslothai/unsloth/pull/12930
* Studio: simpler icons for the audio pages, one audio icon in the Library by @Etherll in https://github.com/unslothai/unsloth/pull/12890
* Desktop contract: count #12927's scaled sidebar row by @danielhanchen in https://github.com/unslothai/unsloth/pull/12931
* fix(studio): preserve skill mention intent and denied preload context by @wasimysaid in https://github.com/unslothai/unsloth/pull/12841
* Studio: embedding model picker, pins and eject in the RAG menu by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/12875
* install-kernels: skip mamba_ssm below sm80 by @danielhanchen in https://github.com/unslothai/unsloth/pull/12921
* Gemma-4 26B/31B: train with the empty thought channel on non-thinking turns by @danielhanchen in https://github.com/unslothai/unsloth/pull/12867
* Studio: try a decision from the Decision API settings by @NilayYadav in https://github.com/unslothai/unsloth/pull/12916
* Studio: confirm browser downloads, choose the download folder by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/12832
* Studio: default Qwen-Image-2.1 int8 to the hosted ConvRot file, with shared rotations by @danielhanchen in https://github.com/unslothai/unsloth/pull/12874
* Studio: recognise ComfyUI checkpoints by name and header, and list ComfyUI model folders by @danielhanchen in https://github.com/unslothai/unsloth/pull/12878
* studio: install the gstreamer recording plugins with the deb package by @mahiatlinux in https://github.com/unslothai/unsloth/pull/12905
* Studio: turn [1]-style citations in replies into links by @NilayYadav in https://github.com/unslothai/unsloth/pull/12912
* Studio: include the chat's system prompt in exports and training data by @NilayYadav in https://github.com/unslothai/unsloth/pull/12913
* studio: guard project submits during IME composition by @mahiatlinux in https://github.com/unslothai/unsloth/pull/12924
* studio: scope speech download cancellation to its attempt by @mahiatlinux in https://github.com/unslothai/unsloth/pull/12925
* studio: keep general settings from restoring stale tokens by @mahiatlinux in https://github.com/unslothai/unsloth/pull/12922
* Studio: show when Windows MXC already runs in the built-in container by @danielhanchen in https://github.com/unslothai/unsloth/pull/12938
* Studio: key the diffusion compile cache by the loaded quant variant by @danielhanchen in https://github.com/unslothai/unsloth/pull/12887
* Studio: rebuild an image / video GGUF from the cached copy when only its header changed by @danielhanchen in https://github.com/unslothai/unsloth/pull/12928
* Scope UNSLOTH_HIGH_PRECISION_LAYERNORM to the load that sets it by @danielhanchen in https://github.com/unslothai/unsloth/pull/12873
* fix(studio): honor llama.cpp update dismissal and snooze by @wasimysaid in https://github.com/unslothai/unsloth/pull/12934
* Keep flash attention from reading Qwen3.5 mRoPE position ids as packed sequences by @danielhanchen in https://github.com/unslothai/unsloth/pull/12856
* Studio: load Wan2.2-A14B expert pairs and tell LTX-2.3 distilled from dev by its weights by @danielhanchen in https://github.com/unslothai/unsloth/pull/12872
* Train a decision model from a plain language model by @danielhanchen in https://github.com/unslothai/unsloth/pull/12772
* Train any text or vision LLM as a Clef decision model from Studio, with FastDecisionModel.predict and adapter saves by @danielhanchen in https://github.com/unslothai/unsloth/pull/12876
* Studio: install transformers releases needing hub >= 1.31, and transformers main after consent by @danielhanchen in https://github.com/unslothai/unsloth/pull/12871
* Correct sample packing for hybrid models (gated-delta, Mamba2, short conv) by @kfastino in https://github.com/unslothai/unsloth/pull/9812
* Studio: read replies aloud without markdown symbols by @AzizMuminov in https://github.com/unslothai/unsloth/pull/12598
* Studio: finish an update with the setup script it installed by @Etherll in https://github.com/unslothai/unsloth/pull/12897
* Serve decision models through llama.cpp in Studio, and export them to GGUF by @danielhanchen in https://github.com/unslothai/unsloth/pull/12939
* studio: fix live monitor background in light mode by @mahiatlinux in https://github.com/unslothai/unsloth/pull/12904
* Studio: show the hosted text encoder download in image load progress by @Etherll in https://github.com/unslothai/unsloth/pull/12894
* Studio: update the audio.cpp runtime from the in-app update by @Etherll in https://github.com/unslothai/unsloth/pull/12893
* Settings contract: read the embedding picker's stacking classes inside cn() too by @danielhanchen in https://github.com/unslothai/unsloth/pull/12945
* install-kernels: install mamba_ssm on sm75 with Triton 3.4+ by @danielhanchen in https://github.com/unslothai/unsloth/pull/12944
* Studio: load Wan2.2 hosted FP8 / INT8 files, and hosted files under low_vram by @danielhanchen in https://github.com/unslothai/unsloth/pull/12888
* Studio: load a single .safetensors DiT from any Hugging Face repo by @danielhanchen in https://github.com/unslothai/unsloth/pull/12879
* fix(studio): remember per-GPU layer ratios by @Imagineer99 in https://github.com/unslothai/unsloth/pull/12774
* Studio: load ComfyUI Krea-2 and HunyuanImage-2.1 single-file DiTs by @danielhanchen in https://github.com/unslothai/unsloth/pull/12885
* Studio: load ComfyUI text encoder and VAE files beside a single-file DiT by @danielhanchen in https://github.com/unslothai/unsloth/pull/12883
* Studio: load ComfyUI nvfp4 and mxfp8 DiT single files by @danielhanchen in https://github.com/unslothai/unsloth/pull/12877
* Studio: keep $PATH, $HOME and other shell variables as text, not maths by @NilayYadav in https://github.com/unslothai/unsloth/pull/12911
* Studio: price the context meter off a tool loop's final pass, not the whole turn's completions by @sumingwang233 in https://github.com/unslothai/unsloth/pull/12889
* Studio: open Try a decision with the trained model after Use in Decision API by @NilayYadav in https://github.com/unslothai/unsloth/pull/12953
* Studio: harden browser downloads after #12832 by @danielhanchen in https://github.com/unslothai/unsloth/pull/12943
* Repair four checks that went red on main with the decision-model merges by @danielhanchen in https://github.com/unslothai/unsloth/pull/12954
* Bump install.sh / install.ps1 pins to unsloth>=2026.10.2 by @danielhanchen in https://github.com/unslothai/unsloth/pull/12960
* install-kernels: tidy the sm75 mamba_ssm follow-ups by @danielhanchen in https://github.com/unslothai/unsloth/pull/12948
* Studio frontend: bump proxy-addr, seroval and MCP SDK for npm advisories by @danielhanchen in https://github.com/unslothai/unsloth/pull/12932
* Studio: close managed-account and API-key gaps in owner-only routes by @danielhanchen in https://github.com/unslothai/unsloth/pull/12940
* Studio: harden S3 dataset keys, uv fallback, header reads and auth body cap by @danielhanchen in https://github.com/unslothai/unsloth/pull/12936
* Decision models: follow-up fixes after #12772, #12876, #12939 by @danielhanchen in https://github.com/unslothai/unsloth/pull/12949
* Fix the backend CI guards main fails after the decision and ComfyUI merges by @danielhanchen in https://github.com/unslothai/unsloth/pull/12958
* Baseline the two unsloth-zoo 2026.10.2 findings after review by @danielhanchen in https://github.com/unslothai/unsloth/pull/12956
* Installer differential: ignore winget spinner frames in the transcript by @danielhanchen in https://github.com/unslothai/unsloth/pull/12965
* tests: keep the Kaggle launcher's signal handlers out of the pytest worker by @danielhanchen in https://github.com/unslothai/unsloth/pull/12974
* Stop test_dataset_cache_safe leaking the Hub no-symlink switch into later tests by @danielhanchen in https://github.com/unslothai/unsloth/pull/12973
* Studio: keep Deep Research tables intact when a cited title has a pipe by @NilayYadav in https://github.com/unslothai/unsloth/pull/12990
* Studio: keep tool call arguments in Qwen3.5 safetensors and MLX prompts by @NilayYadav in https://github.com/unslothai/unsloth/pull/12988
* Fix linked-folder indexing of hidden subdirectories by @Imagineer99 in https://github.com/unslothai/unsloth/pull/12972
* Studio: compact long chats on self-hosted connections to the window the server reports by @oobabooga in https://github.com/unslothai/unsloth/pull/12975
* Studio: show project sources as unused on models without tools by @NilayYadav in https://github.com/unslothai/unsloth/pull/12993
* Studio: show a download card when the python tool edits an attached file by @NilayYadav in https://github.com/unslothai/unsloth/pull/12992
* Studio: stop crashing on a lowercase boolean in PYTORCH_ALLOC_CONF by @oobabooga in https://github.com/unslothai/unsloth/pull/12976
* Studio: decode pages in the browser panel the way browsers do by @NilayYadav in https://github.com/unslothai/unsloth/pull/12983
* Studio: use llama-server for embedding models the installed sentence-transformers cannot load by @oobabooga in https://github.com/unslothai/unsloth/pull/13005
* Studio: train vision datasets that have some rows without an image by @NilayYadav in https://github.com/unslothai/unsloth/pull/12991
* Support text attachments in per-chat prompt queues by @Imagineer99 in https://github.com/unslothai/unsloth/pull/12964
* Studio: make Thinking off and Preserve thinking work on Qwen3.6 by @NilayYadav in https://github.com/unslothai/unsloth/pull/12989
* Studio: smaller settings info icons, engine notes under the engine name by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/13008
* Studio: static dark dropdown glow, and stop modal opens restyling the page by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/13013
* Studio: open video attachments in the browser panel by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/13007
* Studio: show a site's own error page in the browser panel by @NilayYadav in https://github.com/unslothai/unsloth/pull/12986
* Studio: list typed-decisions datasets first for decision training by @NilayYadav in https://github.com/unslothai/unsloth/pull/12982
* Keep the model a training script saves when it runs in Docker by @NilayYadav in https://github.com/unslothai/unsloth/pull/12994
* Skip the fast LoRA paths when lora_B has a bias by @vineethsaivs in https://github.com/unslothai/unsloth/pull/12981
* Studio: prefer llama-server for EmbeddingGemma in the embedding model picker by @oobabooga in https://github.com/unslothai/unsloth/pull/13006
* Studio: ask before numpy pickle loads, keep audio tags inline, cap the variable-prose regex by @danielhanchen in https://github.com/unslothai/unsloth/pull/13001
* Studio: parse Yahoo's newer result layout in web search by @oobabooga in https://github.com/unslothai/unsloth/pull/12980
* Update the flex large head dim mask tests to the unpadded causal-mask skip by @danielhanchen in https://github.com/unslothai/unsloth/pull/13018
* Studio: serve whisper-server under a random per-launch request path by @danielhanchen in https://github.com/unslothai/unsloth/pull/13002
* Studio: find nvidia-smi under WSL on every query, count Core Ultra Arc iGPUs as XPU, cut GPU masks at an invalid index by @danielhanchen in https://github.com/unslothai/unsloth/pull/12962
* Studio: make llama-fit-params executable after the macOS prebuilt install by @jayzhou2309 in https://github.com/unslothai/unsloth/pull/12917
* Studio: leave a bracketed IPv6 host unchanged in dial_host by @drakeo338 in https://github.com/unslothai/unsloth/pull/12733
* Studio: build VAE tile blend weights on CPU so MPS tiled encode works by @gokay-ai in https://github.com/unslothai/unsloth/pull/12937
* Studio: run llama-server with a per-launch API key by default by @danielhanchen in https://github.com/unslothai/unsloth/pull/13011
* Studio: keep new image sets from merging into an existing one by @NilayYadav in https://github.com/unslothai/unsloth/pull/12987
* install.sh: install for the discrete AMD GPU when an iGPU is listed first by @danielhanchen in https://github.com/unslothai/unsloth/pull/12963
* FastModel: fall back to Unsloth inference on GPUs older than Volta by @danielhanchen in https://github.com/unslothai/unsloth/pull/12959
* Unsloth Studio (AMD): show each GPU's live VRAM and utilization under its own HIP id by @danielhanchen in https://github.com/unslothai/unsloth/pull/12961
* Studio: Downloads button in the browser panel by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/13009
* Studio: keep a chat's HTML pages with that chat by @NilayYadav in https://github.com/unslothai/unsloth/pull/12985
* Studio: make the browser's right-click downloads work on macOS by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/13003
* Studio: put the skill row chevron next to the skill name by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/13033
* Studio: close refused tool calls under tool_choice none by @Beverly621 in https://github.com/unslothai/unsloth/pull/12627
* Studio: keep browsing from a temporary chat out of browser history by @NilayYadav in https://github.com/unslothai/unsloth/pull/12984
* Studio: duplicate a past training run into a new Configure draft by @Padi142 in https://github.com/unslothai/unsloth/pull/12979
* Unsloth Studio / Desktop: show a cached image GGUF as Partial until Run has its text encoder and VAE by @LeoBorcherding in https://github.com/unslothai/unsloth/pull/12557
* Load one tensor at a time when quantizing a 16-bit checkpoint, so Qwen3.5-27B 4-bit loads without Block Swap by @LeoBorcherding in https://github.com/unslothai/unsloth/pull/12997
* Studio: show the real cause when a desktop install runs out of disk space by @huntersgordon in https://github.com/unslothai/unsloth/pull/12919
* Fix notebook failures on Kaggle T4x2 and duplicate import warnings by @danielhanchen in https://github.com/unslothai/unsloth/pull/13024
* Studio: skip the Xet probe's GPU-init-off zoo retry on GPU hosts by @arcusbuilds in https://github.com/unslothai/unsloth/pull/12478
* Studio: stop MXC read grants looping on a Microsoft Store Python by @danielhanchen in https://github.com/unslothai/unsloth/pull/13025
* Studio: avoid a TypeError after a tools module reload by @jiangLLM in https://github.com/unslothai/unsloth/pull/11506
* Studio: retry web search over HTTP/1.1 when the connection is reset by @danielhanchen in https://github.com/unslothai/unsloth/pull/13032
* Studio: stop a Git Bash nul file from blocking every Windows MXC tool call by @danielhanchen in https://github.com/unslothai/unsloth/pull/13037
* Decline the packed INT4 kernel when weight_scale has the wrong group count by @jayzhou2309 in https://github.com/unslothai/unsloth/pull/12957
* Studio: let another account join the resident GGUF instead of replacing it by @danielhanchen in https://github.com/unslothai/unsloth/pull/13038
* Studio: give the annotate comment box a visible shadow in dark mode by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/13071
* Installer: read the venv's torch in isolation so a PYTHONPATH torch cannot break the install by @danielhanchen in https://github.com/unslothai/unsloth/pull/13041
* Studio: keep tensor split and MTP when Auto would pick a DFlash drafter that aborts it by @danielhanchen in https://github.com/unslothai/unsloth/pull/13040
* Allow backward through eval-mode and for_inference forwards by @danielhanchen in https://github.com/unslothai/unsloth/pull/13050
* Make Q-GaLore optimizer state resumable from checkpoints by @danielhanchen in https://github.com/unslothai/unsloth/pull/13051
* Studio: Docs links for Sandbox, Agents and the Decision API by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/13076
* Studio: serve decision models through the MLX engine on Apple Silicon by @Lyxot in https://github.com/unslothai/unsloth/pull/13015
* Studio: pin FastFlowLM 1.0.7 for Qwen3.8 27B on the AMD NPU and keep Lemonade / FastFlowLM current by @danielhanchen in https://github.com/unslothai/unsloth/pull/13047
* Studio: report the real llama.cpp version for source builds by @danielhanchen in https://github.com/unslothai/unsloth/pull/13036
* Studio: name media companion downloads by component, and show what Run still fetches for an on-device GGUF by @danielhanchen in https://github.com/unslothai/unsloth/pull/13027
* Studio: keep a dragged annotate area the size it was drawn by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/13070
* Studio: load a repo id from its scan-folder copy instead of re-downloading it by @danielhanchen in https://github.com/unslothai/unsloth/pull/13034
* Train EXAONE 3.5: name the token embedding remote code no longer exposes by @danielhanchen in https://github.com/unslothai/unsloth/pull/13058

## New Contributors
* @kfastino made their first contribution in https://github.com/unslothai/unsloth/pull/9812
* @sumingwang233 made their first contribution in https://github.com/unslothai/unsloth/pull/12889
* @drakeo338 made their first contribution in https://github.com/unslothai/unsloth/pull/12733
* @Beverly621 made their first contribution in https://github.com/unslothai/unsloth/pull/12627
* @Padi142 made their first contribution in https://github.com/unslothai/unsloth/pull/12979
* @huntersgordon made their first contribution in https://github.com/unslothai/unsloth/pull/12919
* @arcusbuilds made their first contribution in https://github.com/unslothai/unsloth/pull/12478
* @jiangLLM made their first contribution in https://github.com/unslothai/unsloth/pull/11506

**Full Changelog**: https://github.com/unslothai/unsloth/compare/v0.1.903-beta...v0.1.905-beta