Skip to content

Unsloth

v0.1.43-beta Security

This release includes 1 security fix for security teams reviewing exposed deployments.

Published 3d Model Serving & MLOps
✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →
This release patches 1 known CVE

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

deps breaking_upgrade

ReleasePort's take

Moderate signal
editorial:auto 3d

Bump openssl, rand, rustls-webpki, and tar in Studio src‑tauri to address security advisories. Add --parallel/-np flag on `unsloth studio run` for parallel execution.

Why it matters: Security advisory bumps (severity 80) mitigate vulnerabilities; the new --parallel flag (severity 50) enables up to N‑times faster runs depending on CPU cores.

Summary

AI summary

Broad release touches Blackwell GPUs Update, Studio, Mac Updates, and Windows Updates.

Changes in this release

Security High

Bump openssl, rand, rustls-webpki, tar in Studio src-tauri for security advisories.

Bump openssl, rand, rustls-webpki, tar in Studio src-tauri for security advisories.

Source: llm_adapter@2026-05-31

Confidence: high

 —
Breaking High

Do not use `unsloth studio update`; it does not fetch latest updates.

Do not use `unsloth studio update`; it does not fetch latest updates.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Feature Medium

Expose --parallel / -np flag on `unsloth studio run`.

Expose --parallel / -np flag on `unsloth studio run`.

Source: llm_adapter@2026-05-31

Confidence: high

 —
Feature Medium

Add Gemini provider with web_search, code_execution, prompt caching, and Nano Banana image generation.

Add Gemini provider with web_search, code_execution, prompt caching, and Nano Banana image generation.

Source: llm_adapter@2026-05-31

Confidence: high

 —
Feature Medium

Add configurable CPU thread pool limit in Studio.

Add configurable CPU thread pool limit in Studio.

Source: llm_adapter@2026-05-31

Confidence: high

 —
Feature Medium

Expose image size setting in Studio training UI.

Expose image size setting in Studio training UI.

Source: llm_adapter@2026-05-31

Confidence: high

 —
Feature Medium

Add remote MCP server support in Studio.

Add remote MCP server support in Studio.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Feature Medium

Keep web search/code pills off on model load if user disabled them in Studio.

Keep web search/code pills off on model load if user disabled them in Studio.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Feature Medium

Add frontend i18n support in Studio.

Add frontend i18n support in Studio.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Feature Medium

Add tool mask support.

Add tool mask support.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Feature Low

Add stdio MCP server support in Studio.

Add stdio MCP server support in Studio.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Performance Medium

Harden Linux llama.cpp prebuilt reuse.

Harden Linux llama.cpp prebuilt reuse.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Performance Medium

Make NVIDIA prebuilt selection track CUDA version bumps (Windows + Linux).

Make NVIDIA prebuilt selection track CUDA version bumps (Windows + Linux).

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Performance Medium

Pick a macOS llama.cpp prebuilt that loads on the host OS in Studio.

Pick a macOS llama.cpp prebuilt that loads on the host OS in Studio.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Performance Medium

Pin the last pre‑macOS‑26 llama.cpp prebuilt instead of walking back in Studio setup.

Pin the last pre‑macOS‑26 llama.cpp prebuilt instead of walking back in Studio setup.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Bugfix Medium

Fix non‑streaming GGUF chat completion usage.

Fix non‑streaming GGUF chat completion usage.

Source: llm_adapter@2026-05-31

Confidence: high

 —
Bugfix Medium

Fix error when context‑resizing.

Fix error when context‑resizing.

Source: llm_adapter@2026-05-31

Confidence: high

 —
Bugfix Medium

Fix Unsloth Studio not loading in Colab (iframe refused to connect and wrong URL).

Fix Unsloth Studio not loading in Colab (iframe refused to connect and wrong URL).

Source: llm_adapter@2026-05-31

Confidence: high

 —
Bugfix Medium

Detect CUDA UMD Version from newer nvidia‑smi output (fixes #5812).

Detect CUDA UMD Version from newer nvidia‑smi output (fixes #5812).

Source: llm_adapter@2026-05-31

Confidence: high

 —
Bugfix Medium

Honor --ctx-size and other forwarded args from `unsloth studio run` in Studio's context‑fit logic.

Honor --ctx-size and other forwarded args from `unsloth studio run` in Studio's context‑fit logic.

Source: llm_adapter@2026-05-31

Confidence: high

 —
Bugfix Medium

Fix phantom scroll area below messages with many sources in Studio.

Fix phantom scroll area below messages with many sources in Studio.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Bugfix Low

Fix MCP dialog overflow on long URLs in Studio frontend.

Fix MCP dialog overflow on long URLs in Studio frontend.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Refactor Low

Remove dead top-level bitsandbytes import from models/_utils.py.

Remove dead top-level bitsandbytes import from models/_utils.py.

Source: granite4.1:30b@2026-05-31-audit

Confidence: low

 —
Full changelog

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

Mac Updates

  • Re-enabled llama.cpp prebuilt binaries for Apple Silicon (M1-M4) - Mac OS 14 / 15 / 26 (Tahoe)
  • Apple Silicon Mac OS 13 (Ventura) is source build
  • Intel (x86_64) for Mac OS 13.3 / 14 / 15 / 26 (Tahoe) uses llama.cpp prebuilt binaries
  • Intel for Max 13.0 - 13.2 is source build

Windows Updates

  • CUDA 13.3 llama.cpp prebuilt binaries now work for Windows
  • For CUDA 13.2, CUDA 13.1 and below, Windows devices uses CUDA 12.4 fallback - we'll work on CUDA 13.1 binaries soon.

CUDA 13.3 Update

  • CUDA 13.3 non Linux binaries work. We'll still use CUDA 13.1 for now
  • CUDA 13.3 solves the CUDA 13.2 gibberish problem - see https://github.com/unslothai/unsloth/issues/4849

Blackwell GPUs Update

  • For now Blackwell will have delayed releases of llama.cpp prebuilt binaries sine CUDA 12.4 does not work - we are working to resolve this soon.

What's Changed

  • Bump install.sh / install.ps1 pin to unsloth>=2026.5.8 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5791
  • Studio: expose --parallel / -np flag on unsloth studio run by @danielhanchen in https://github.com/unslothai/unsloth/pull/5737
  • tests: unblock three stale assertions broken on main (MLX CI + Backend CI) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5803
  • ci: install unsloth_zoo from git main in notebooks-ci + studio-backend-ci by @danielhanchen in https://github.com/unslothai/unsloth/pull/5802
  • tool mask support by @Datta0 in https://github.com/unslothai/unsloth/pull/5682
  • Studio: add frontend i18n support by @penumbrazz in https://github.com/unslothai/unsloth/pull/5765
  • Studio: unblock install on Linux ARM64 + Windows ARM64 + Intel Mac by @danielhanchen in https://github.com/unslothai/unsloth/pull/5790
  • Harden Linux llama.cpp prebuilt reuse by @mmathew23 in https://github.com/unslothai/unsloth/pull/5796
  • Studio: expose image size setting in training UI by @Dariton4000 in https://github.com/unslothai/unsloth/pull/5743
  • Studio: add configurable CPU thread pool limit by @alkinun in https://github.com/unslothai/unsloth/pull/5760
  • Trim README Advanced launch options blurb by @danielhanchen in https://github.com/unslothai/unsloth/pull/5809
  • [RL] make sync weights conditional by @Datta0 in https://github.com/unslothai/unsloth/pull/4925
  • Studio: add Gemini provider with web_search, code_execution, prompt caching, and Nano Banana image generation by @danielhanchen in https://github.com/unslothai/unsloth/pull/5720
  • Studio: add remote MCP server support by @NilayYadav in https://github.com/unslothai/unsloth/pull/5750
  • Clear MRoPE after generation for GRPO by @Datta0 in https://github.com/unslothai/unsloth/pull/5683
  • Detect CUDA UMD Version from newer nvidia-smi output (fixes #5812) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5817
  • Studio: remove dark mode upload circle by @Imagineer99 in https://github.com/unslothai/unsloth/pull/5813
  • fix: honor --ctx-size and other forwarded args from unsloth studio run in Studio's context-fit logic by @mvanhorn in https://github.com/unslothai/unsloth/pull/5815
  • Fix non-streaming GGUF chat completion usage by @alkinun in https://github.com/unslothai/unsloth/pull/5781
  • Fix/studio colab proxy and iframe - Unsloth Studio not loading in Colab (iframe "refused to connect" and wrong URL) by @LeoBorcherding in https://github.com/unslothai/unsloth/pull/5844
  • Studio: keep web search/code pills off on model load if user disabled them by @rolandtannous in https://github.com/unslothai/unsloth/pull/5851
  • studio/setup.sh: cope with fresh CUDA toolkits like 13.3 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5826
  • fix/strix halo and windows AMD ROCm support by @LeoBorcherding in https://github.com/unslothai/unsloth/pull/5301
  • Version downpairing for Radeon+ROCm PyTorch wheels by @jaeiclee in https://github.com/unslothai/unsloth/pull/5353
  • feature/use lemonade-sdk llamacpp-rocm binaries by @LeoBorcherding in https://github.com/unslothai/unsloth/pull/5303
  • studio: ROCm cleanups follow-up to #5301 by @danielhanchen in https://github.com/unslothai/unsloth/pull/5874
  • fix: wrong code block language for nightly Windows install by @mervivian in https://github.com/unslothai/unsloth/pull/5877
  • fix: bump openssl, rand, rustls-webpki, tar in Studio src-tauri for security advisories by @orbisai0security in https://github.com/unslothai/unsloth/pull/5866
  • Fix error when context-resizing by @afsuyadi in https://github.com/unslothai/unsloth/pull/5860
  • Update Studio sidebar logout icon to logout-05 by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/5878
  • studio: pick a macOS llama.cpp prebuilt that loads on the host OS by @danielhanchen in https://github.com/unslothai/unsloth/pull/5883
  • studio: make NVIDIA prebuilt selection track CUDA version bumps (Windows + Linux) by @danielhanchen in https://github.com/unslothai/unsloth/pull/5879
  • Remove dead top-level bitsandbytes import from models/_utils.py by @Rahul007007 in https://github.com/unslothai/unsloth/pull/5875
  • ci(security-audit): make package installs network-resilient by @danielhanchen in https://github.com/unslothai/unsloth/pull/5853
  • Studio: add stdio MCP server support by @oobabooga in https://github.com/unslothai/unsloth/pull/5863
  • studio/frontend: fix MCP dialog overflow on long URLs by @oobabooga in https://github.com/unslothai/unsloth/pull/5864
  • Studio: clearer error for diffusion GGUFs loaded as chat models by @danielhanchen in https://github.com/unslothai/unsloth/pull/5857
  • Studio: harden stdio MCP gating and fix transport edge cases by @danielhanchen in https://github.com/unslothai/unsloth/pull/5892
  • Studio: fix phantom scroll area below messages with many sources by @danielhanchen in https://github.com/unslothai/unsloth/pull/5822
  • Studio: pin the last pre-macOS-26 llama.cpp prebuilt instead of walking back by @danielhanchen in https://github.com/unslothai/unsloth/pull/5896
  • fix(install): detect x86_64 Python venv on Apple Silicon and rebuild as arm64 by @ramankrishna in https://github.com/unslothai/unsloth/pull/5187

New Contributors

  • @penumbrazz made their first contribution in https://github.com/unslothai/unsloth/pull/5765
  • @Dariton4000 made their first contribution in https://github.com/unslothai/unsloth/pull/5743
  • @NilayYadav made their first contribution in https://github.com/unslothai/unsloth/pull/5750
  • @mvanhorn made their first contribution in https://github.com/unslothai/unsloth/pull/5815
  • @jaeiclee made their first contribution in https://github.com/unslothai/unsloth/pull/5353
  • @mervivian made their first contribution in https://github.com/unslothai/unsloth/pull/5877
  • @orbisai0security made their first contribution in https://github.com/unslothai/unsloth/pull/5866
  • @afsuyadi made their first contribution in https://github.com/unslothai/unsloth/pull/5860
  • @Rahul007007 made their first contribution in https://github.com/unslothai/unsloth/pull/5875
  • @ramankrishna made their first contribution in https://github.com/unslothai/unsloth/pull/5187

Full Changelog: https://github.com/unslothai/unsloth/compare/v0.1.42-beta...v0.1.43-beta

Security Fixes

  • Bump openssl, rand, rustls-webpki, tar in Studio src-tauri for security advisories
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

Earlier breaking changes

  • 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]