Skip to content

Unsloth

v0.1.45-beta Feature

This release adds 3 notable features for engineering teams evaluating rollout.

Published 1mo Model Serving & MLOps
✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

agent deepseek fine-tuning gemma gemma3 gpt-oss
+13 more
llama llama3 llm llms mistral openai qwen reinforcement-learning self-hosted text-to-speech tts ui unsloth

Affected surfaces

breaking_upgrade

ReleasePort's take

Moderate signal
editorial:auto 1mo

The release blocks arbitrary external image URLs in Studio's markdown renderer and adds an in‑app button to update llama.cpp prebuilt binaries.

Why it matters: Blocking unsafe image URLs prevents remote content injection; the new Update button simplifies binary maintenance for Windows/WSL installs.

Summary

AI summary

Studio updates, CI changes, installer fixes, GPU/ROCm handling, UI polish across modules.

Changes in this release

Security High

Blocks arbitrary external image URLs in markdown renderer.

Blocks arbitrary external image URLs in markdown renderer.

Source: llm_adapter@2026-06-10

Confidence: high

 —
Feature Medium

Adds Hub + Download Manager to Studio.

Adds Hub + Download Manager to Studio.

Source: llm_adapter@2026-06-10

Confidence: high

 —
Feature Medium

Enables audio input for Gemma 4 GGUF models in Studio.

Enables audio input for Gemma 4 GGUF models in Studio.

Source: llm_adapter@2026-06-10

Confidence: high

 —
Feature Medium

Sets Qwen3.5‑4B‑MTP as the default chat model in Studio.

Sets Qwen3.5‑4B‑MTP as the default chat model in Studio.

Source: llm_adapter@2026-06-10

Confidence: high

 —
Dependency Medium

Sources llama.cpp prebuilt binaries from unslothai/llama.cpp repository.

Sources llama.cpp prebuilt binaries from unslothai/llama.cpp repository.

Source: llm_adapter@2026-06-10

Confidence: high

 —
Bugfix High

Prevents mid‑stream refresh from losing the prompt in Studio.

Prevents mid‑stream refresh from losing the prompt in Studio.

Source: llm_adapter@2026-06-10

Confidence: low

 —
Bugfix Medium

Fixes Gemma‑4‑12B‑it model loading issue in Studio.

Fixes Gemma‑4‑12B‑it model loading issue in Studio.

Source: llm_adapter@2026-06-10

Confidence: high

 —
Bugfix Medium

Improves Windows/WSL installer handling of winget cert failures, DiskPart prompts, and enables AMD GPU support.

Improves Windows/WSL installer handling of winget cert failures, DiskPart prompts, and enables AMD GPU support.

Source: llm_adapter@2026-06-10

Confidence: high

 —
Bugfix Medium

Provides clearer setup error when GPU driver is too old for the installed CUDA toolkit.

Provides clearer setup error when GPU driver is too old for the installed CUDA toolkit.

Source: llm_adapter@2026-06-10

Confidence: high

 —
Bugfix Medium

Adds an in‑app button to update llama.cpp prebuilt binaries.

Adds an in‑app button to update llama.cpp prebuilt binaries.

Source: llm_adapter@2026-06-10

Confidence: low

 —
Bugfix Medium

Persists Studio thread synchronously on first runStart to keep prompt during mid‑stream refresh.

Persists Studio thread synchronously on first runStart to keep prompt during mid‑stream refresh.

Source: granite4.1:30b@2026-06-10-audit

Confidence: low

 —
Full changelog

Bug Fixes and more cross platform support

To update Unsloth or install a new Unsloth Studio, you must use:

macOS, Linux, WSL:

curl -fsSL https://unsloth.ai/install.sh | sh

Windows:

irm https://unsloth.ai/install.ps1 | iex

[!WARNING]
DO NOT USE unsloth studio update since packaging will not get the latest updates

What's Changed

  • Bump install.sh / install.ps1 pin to unsloth>=2026.6.1 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5977
  • Port KTO logps truncation guard to TRL 1.x _compute_logps refactor by @danielhanchen in https://github.com/unslothai/unsloth/pull/5996
  • CI: track deepseek_ocr2 compile timeout in known-broken list by @danielhanchen in https://github.com/unslothai/unsloth/pull/5995
  • fix(studio): disable mlx gc for none by @Lyxot in https://github.com/unslothai/unsloth/pull/5991
  • Normalize shell scripts to LF in .gitattributes by @danielhanchen in https://github.com/unslothai/unsloth/pull/5997
  • Studio: enable audio input for Gemma 4 GGUFs; default chat model to Qwen3.5-4B-MTP by @danielhanchen in https://github.com/unslothai/unsloth/pull/6000
  • Fix chat text cutoff at composer dock and speed up plus icon spin by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5989
  • Studio: refine tool call and reasoning trigger UI by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5873
  • fix: warn when localhost resolves to ::1 but Studio is bound only to 127.0.0.1 by @mvanhorn in https://github.com/unslothai/unsloth/pull/5994
  • fix: persist Studio thread synchronously on first runStart so mid-stream refresh keeps the prompt by @mvanhorn in https://github.com/unslothai/unsloth/pull/5814
  • Studio: enable GGUF tools with vision inputs by @Imagineer99 in https://github.com/unslothai/unsloth/pull/6009
  • Studio: accept system-role messages in Claude Code requests by @Imagineer99 in https://github.com/unslothai/unsloth/pull/6006
  • Studio: fix load_freeze audio-type tests for #6000's Gemma 4 <|audio|> probe by @danielhanchen in https://github.com/unslothai/unsloth/pull/6018
  • Studio: fix chat preset persistence with fast mode by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5870
  • Studio: fix Repo tests (CPU) by stopping the ROCm test from leaking a fake utils into sys.modules by @danielhanchen in https://github.com/unslothai/unsloth/pull/6027
  • Studio: stop ROCm amd-smi tests leaking a fake loggers into sys.modules (follow-up to #6027) by @danielhanchen in https://github.com/unslothai/unsloth/pull/6055
  • Studio: emit usage and timings for MLX generation speed stats by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6068
  • Studio: tag MLX loaded models as MLX instead of Base in chat by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6067
  • Studio: remove red border on chat error messages by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6063
  • Studio: keep chat in place when composer attachments resize it by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6070
  • CI: allowlist deepseek_ocr2 in compiler full-model-sweep by @danielhanchen in https://github.com/unslothai/unsloth/pull/6085
  • Restore KTO logps truncation guard for TRL (re-apply dropped #5996) by @danielhanchen in https://github.com/unslothai/unsloth/pull/6086
  • Studio: stop leaking internal exceptions to API clients; harden sandbox path by @danielhanchen in https://github.com/unslothai/unsloth/pull/6072
  • Formatting: ruff line-length 100 + drop blank after short local imports by @danielhanchen in https://github.com/unslothai/unsloth/pull/6079
  • Fix MoE LoRA target parameter handling by @Datta0 in https://github.com/unslothai/unsloth/pull/5345
  • Refactor VLM detection in studio by @Datta0 in https://github.com/unslothai/unsloth/pull/5245
  • qwen 3.5 export fixes by @Datta0 in https://github.com/unslothai/unsloth/pull/5992
  • [fix] Nvfp4 load by @Datta0 in https://github.com/unslothai/unsloth/pull/6087
  • Studio: open the MCP dialog to the server list by @oobabooga in https://github.com/unslothai/unsloth/pull/6100
  • Studio: make code comments and docstrings more succinct by @danielhanchen in https://github.com/unslothai/unsloth/pull/6029
  • Reduce and tighten code comments and docstrings repo-wide by @danielhanchen in https://github.com/unslothai/unsloth/pull/6095
  • Studio frontend: reduce and tighten code comments by @danielhanchen in https://github.com/unslothai/unsloth/pull/6099
  • feat(studio): Hub + Download Manager by @Sneakr in https://github.com/unslothai/unsloth/pull/5916
  • Studio fix recipe dataset preview by @wasimysaid in https://github.com/unslothai/unsloth/pull/6031
  • Studio: make Helper LLM startup pre-cache opt in by @wasimysaid in https://github.com/unslothai/unsloth/pull/6113
  • Improve local chat tool call flow by @wasimysaid in https://github.com/unslothai/unsloth/pull/5962
  • [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in https://github.com/unslothai/unsloth/pull/6104
  • Studio: improve OpenAI- and Anthropic-compatible API spec compliance by @oobabooga in https://github.com/unslothai/unsloth/pull/6010
  • Studio: follow-up fix for GGUF developer prompts by @wasimysaid in https://github.com/unslothai/unsloth/pull/6115
  • Studio: stop the providers dialog from resetting custom provider form state by @NilayYadav in https://github.com/unslothai/unsloth/pull/6051
  • Studio: clean-room compact RAG (knowledge bases, hybrid search, fast indexing) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5910
  • feat: support text-only loading of Gemma 3 27B via FastLanguageModel (skip SiglipVisionModel) by @mvanhorn in https://github.com/unslothai/unsloth/pull/5816
  • studio(ui): use the --primary brand token for the avatar fallback color by @danielhanchen in https://github.com/unslothai/unsloth/pull/5987
  • fix(studio): block arbitrary external image URLs in markdown renderer by @rsd-darshan in https://github.com/unslothai/unsloth/pull/5602
  • Studio: center account avatar vertically in sidebar footer pill by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6026
  • fix: clearer Studio setup error when GPU driver is too old for the installed CUDA toolkit by @mvanhorn in https://github.com/unslothai/unsloth/pull/5993
  • Studio: npm v12 readiness (allowScripts policy, npmrc cleanup, bun bootstrap fix) by @danielhanchen in https://github.com/unslothai/unsloth/pull/6128
  • Studio: faithful conversation export and import round trips (ShareGPT system role, CSV quoted newlines) by @danielhanchen in https://github.com/unslothai/unsloth/pull/6131
  • Studio: auto-sync allowScripts pins after dependency bumps by @danielhanchen in https://github.com/unslothai/unsloth/pull/6136
  • Studio: unify shadows, backgrounds and dark mode consistency in chat UI by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6116
  • Windows/WSL installer: fix winget msstore cert failure, amd-smi DiskPart prompt, and enable AMD GPU (Strix Halo gfx1151) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5940
  • Studio: bulk export and import in Settings Chat Data, MCP pill off switch by @danielhanchen in https://github.com/unslothai/unsloth/pull/6141
  • Studio: fix nested dropdown submenus clipped by the menu alignment nudge by @danielhanchen in https://github.com/unslothai/unsloth/pull/6143
  • Studio: declutter the chat plus menu and make RAG session-only with pre-select by @danielhanchen in https://github.com/unslothai/unsloth/pull/6140
  • fix/validate dataset video paths before training by @LeoBorcherding in https://github.com/unslothai/unsloth/pull/5136
  • Restore config use_cache in for_inference after gradient checkpointing prep by @danielhanchen in https://github.com/unslothai/unsloth/pull/6137
  • fix: persist Windows ROCm BNB version by @Peter7896 in https://github.com/unslothai/unsloth/pull/6048
  • Studio GGUF CI: fix deterministic JSON-mode failure by bumping the quant, keep hard asserts by @danielhanchen in https://github.com/unslothai/unsloth/pull/6138
  • Frontend CI: hard-fail unreviewed npm install scripts (--strict-allow-scripts) by @danielhanchen in https://github.com/unslothai/unsloth/pull/6139
  • Studio UI polish: search dialog shadow, model picker pills, sidebar spacing, white Hub background by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6147
  • Run the HF cache redirect before import fixes that can freeze Hub constants by @danielhanchen in https://github.com/unslothai/unsloth/pull/6150
  • Fix Studio MLX VLM resized image layout by @Lyxot in https://github.com/unslothai/unsloth/pull/6019
  • fix(studio): reject unsupported MLX CPT and embedding training by @Lyxot in https://github.com/unslothai/unsloth/pull/6091
  • fix(studio): forward custom dataset mappings in MLX training by @Lyxot in https://github.com/unslothai/unsloth/pull/6094
  • Tests + CI guard: batched left-padded generation can never silently regress again (#1066, #3699) by @danielhanchen in https://github.com/unslothai/unsloth/pull/6145
  • Tests: follow Compare chat into the More submenu in the Playwright chat UI driver by @danielhanchen in https://github.com/unslothai/unsloth/pull/6153
  • Fix UnboundLocalError in ROCm version detection dpkg/rpm fallback by @danielhanchen in https://github.com/unslothai/unsloth/pull/6149
  • Auto-set BNB_ROCM_VERSION from the installed wheel on Windows + ROCm by @danielhanchen in https://github.com/unslothai/unsloth/pull/5986
  • Call patch_compiling_bitsandbytes on the FastLanguageModel path by @danielhanchen in https://github.com/unslothai/unsloth/pull/6144
  • Studio: mascot images degrade gracefully instead of showing alt text by @danielhanchen in https://github.com/unslothai/unsloth/pull/6146
  • Studio: training survives a non-writable HF datasets cache by @danielhanchen in https://github.com/unslothai/unsloth/pull/6148
  • Studio: fix Gemma-4-12B-it not loading by @Imagineer99 in https://github.com/unslothai/unsloth/pull/6054
  • Installer: never plan a non-sm_120 CUDA build on Blackwell, never plan CPU on an NVIDIA host by @danielhanchen in https://github.com/unslothai/unsloth/pull/6156
  • Studio: support separate-file MTP GGUF drafters (Gemma 4) by @oobabooga in https://github.com/unslothai/unsloth/pull/6125
  • Studio: accept audio files through Add photos & files and fix the audio gate for Gemma 4 models by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6064
  • Source llama.cpp prebuilts from unslothai/llama.cpp (CUDA, ROCm, macOS) by @oobabooga in https://github.com/unslothai/unsloth/pull/5963
  • Studio: fall back to text-only when llama.cpp is too old for a model's vision projector by @danielhanchen in https://github.com/unslothai/unsloth/pull/6075
  • Chat UI polish: template editor reset, circular action button hovers by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/6151
  • feat(hub): enable Run/New Chat for downloaded GGUF models, fix README by @Sneakr in https://github.com/unslothai/unsloth/pull/6152
  • Studio: in-app Update llama.cpp button to install the latest prebuilt by @danielhanchen in https://github.com/unslothai/unsloth/pull/6097
  • Studio: gracefully disable MTP when the model has no head or drafter by @danielhanchen in https://github.com/unslothai/unsloth/pull/6159
  • Studio: show the llama.cpp update banner sooner and keep it until dismissed by @danielhanchen in https://github.com/unslothai/unsloth/pull/6162
  • fix(studio): round compact tooltip corners to 9px by @Sneakr in https://github.com/unslothai/unsloth/pull/6163

New Contributors

  • @Sneakr made their first contribution in https://github.com/unslothai/unsloth/pull/5916
  • @rsd-darshan made their first contribution in https://github.com/unslothai/unsloth/pull/5602
  • @Peter7896 made their first contribution in https://github.com/unslothai/unsloth/pull/6048

Full Changelog: https://github.com/unslothai/unsloth/compare/v0.1.44-beta...v0.1.45-beta

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track Unsloth

Get notified when new releases ship.

Sign up free

About Unsloth

All releases →

Related context

Related tools

Earlier breaking changes

  • v0.1.43-beta Do not use `unsloth studio update`; it does not fetch latest updates.
  • v0.1.40-beta Auto MTP speculative decoding enabled by default for MTP GGUFs; warns on stale llama.cpp prebuilt

Beta — feedback welcome: [email protected]