Meta Muse Glimmer

[danielhanchen]

Meta has released Muse Glimmer, the first open model from Meta Superintelligence Labs. Muse Glimmer is a dense, 30B parameter model by Meta, built for local agentic and coding workflows. Released under the Apache 2.0 license you can run Muse Glimmer 30B via Unsloth and Unsloth Dynamic quants for the best performance.

Run Muse Glimmer
Fine-tune Muse Glimmer

Muse Glimmer 30B can run locally on 20GB RAM/VRAM setups, including Mac and GPUs. You can also train and fine-tune Muse Glimmer 30B using Unsloth on 20GB VRAM.

Updating / installing Unsloth

macOS, Linux, WSL:

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

Windows:

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

What's Changed

• Studio: judge the cached-pipeline signals on the snapshot the row actually loads by @danielhanchen in Studio: judge the cached-pipeline signals on the snapshot the row actually loads #7851
• Bump install.sh / install.ps1 pin to unsloth>=2026.8.3 by @danielhanchen in Bump install.sh / install.ps1 pin to unsloth>=2026.8.3 #7860
• Studio: make the sidebar menu separators visible in dark mode by @shimmyshimmer in Studio: make the sidebar menu separators visible in dark mode #7865
• Studio: run sandbox matplotlib headless by @NilayYadav in Studio: run sandbox matplotlib headless #7789
• Don't send desktop startup requests to an unvalidated port 8888 by @NilayYadav in Don't send desktop startup requests to an unvalidated port 8888 #7852
• Support independent prompt queues across parallel chats by @Imagineer99 in Support independent prompt queues across parallel chats #7629
• add a background image to the macos dmg install window by @mahiatlinux in add a background image to the macos dmg install window #7827
• Studio: keep Video under More until a user pins it by @shimmyshimmer in Studio: keep Video under More until a user pins it #7863
• Studio: show generated media unrounded, and fix the selected thumbnail by @shimmyshimmer in Studio: show generated media unrounded, and fix the selected thumbnail #7883
• Studio: drop the New badge from the Video nav row by @shimmyshimmer in Studio: drop the New badge from the Video nav row #7890
• Studio: stop a clicked tooltip from hanging over dialogs by @shimmyshimmer in Studio: stop a clicked tooltip from hanging over dialogs #7836
• Desktop: move update controls to Settings > General by @oobabooga in Desktop: move update controls to Settings > General #7807
• Fix the stale response-model label assertion that has main red by @danielhanchen in Fix the stale response-model label assertion that has main red #7874
• Studio: detect stalled Deep Research output streams by @alkinun in Studio: detect stalled Deep Research output streams #7839
• Guard the release-desktop version input and document the Cargo version by @danielhanchen in Guard the release-desktop version input and document the Cargo version #7889
• Studio: add Qwen3-ASR dictation, cancellable downloads, and a download prompt by @shimmyshimmer in Studio: add Qwen3-ASR dictation, cancellable downloads, and a download prompt #7835
• Studio: even out the gap around the More flyout's rule by @shimmyshimmer in Studio: even out the gap around the More flyout's rule #7899
• Studio: stop the New chat row looking hovered on a blank new chat by @shimmyshimmer in Studio: stop the New chat row looking hovered on a blank new chat #7898
• Studio: fix remaining translation defects after Studio: improve translations across all supported locales #7797 by @danielhanchen in Studio: fix remaining translation defects after #7797 #7850
• Studio: fix download card alignment and surface the stop button by @shimmyshimmer in Studio: fix download card alignment and surface the stop button #7848
• Keep the file-edit turn 2 prompt to one instruction, and stop waiving its cap by @danielhanchen in Keep the file-edit turn 2 prompt to one instruction, and stop waiving its cap #7846
• Studio: align Tauri sidebar spacing by @wasimysaid in Studio: align Tauri sidebar spacing #7901
• Fix sidebar bottom fade cutting off the last row mid-glyph by @shimmyshimmer in Fix sidebar bottom fade cutting off the last row mid-glyph #7902
• Bump install.sh / install.ps1 pin to unsloth>=2026.8.4 by @danielhanchen in Bump install.sh / install.ps1 pin to unsloth>=2026.8.4 #7906
• Answer GGUF variants from the cache the request names, and skip an unreadable one by @danielhanchen in Answer GGUF variants from the cache the request names, and skip an unreadable one #7857
• Desktop: warn before quitting during an update and skip installer fallback on cancel by @NilayYadav in Desktop: warn before quitting during an update and skip installer fallback on cancel #7793
• Desktop: publish drafts without rebuilding by @wasimysaid in Desktop: publish drafts without rebuilding #7907
• Studio: restore locale parity and gate it in CI by @danielhanchen in Studio: restore locale parity and gate it in CI #7904
• Studio: make training and model-picker copy name UI that exists by @danielhanchen in Studio: make training and model-picker copy name UI that exists #7905
• fix(studio): run bash and name the real platform for the terminal tool on Windows by @LeoBorcherding in fix(studio): run bash and name the real platform for the terminal tool on Windows #7885
• Desktop: fix Linux release asset names by @wasimysaid in Desktop: fix Linux release asset names #7911
• Unsloth Studio: keep the Images and Video headers clear of the custom titlebar by @LeoBorcherding in Unsloth Studio: keep the Images and Video headers clear of the custom titlebar #7918
• fix(studio): keep child spawns alive under a pid-1 parent by @mahiatlinux in fix(studio): keep child spawns alive under a pid-1 parent #7894
• Ask uv for a Python that can import torch, and skip the one that cannot by @danielhanchen in Ask uv for a Python that can import torch, and skip the one that cannot #7830
• Improve desktop startup status by @wasimysaid in Improve desktop startup status #7924
• Tauri smoke: stop uploading the whole target/debug tree (350GB of artifacts) by @danielhanchen in Tauri smoke: stop uploading the whole target/debug tree (350GB of artifacts) #7930
• Restore warnings_issued for trl trainers on transformers 5.1+ by @danielhanchen in Restore warnings_issued for trl trainers on transformers 5.1+ #7854
• Fetch a sentence-transformers model's subfolder weights instead of pruning them by @danielhanchen in Fetch a sentence-transformers model's subfolder weights instead of pruning them #7855
• Do not enter CUDA autocast for a float32 model during generate by @danielhanchen in Do not enter CUDA autocast for a float32 model during generate #7900
• Locate the dictation model search box by test id, not by placeholder copy by @danielhanchen in Locate the dictation model search box by test id, not by placeholder copy #7909
• Record a stopped repair update as canceled rather than failed by @danielhanchen in Record a stopped repair update as canceled rather than failed #7910
• Bound dataset_num_proc by memory, and stop treating 1 as "no multiprocessing" by @danielhanchen in Bound dataset_num_proc by memory, and stop treating 1 as "no multiprocessing" #7831
• Studio: quiet the image and video generation progress polls by @danielhanchen in Studio: quiet the image and video generation progress polls #7869
• Studio: stop the Images Train settings columns overlapping by @danielhanchen in Studio: stop the Images Train settings columns overlapping #7895
• Fake the Windows blocklist in the mangled-switch test so it runs off Windows by @danielhanchen in Fake the Windows blocklist in the mangled-switch test so it runs off Windows #7934
• Refine desktop media header alignment by @wasimysaid in Refine desktop media header alignment #7859
• Read the GGUF context length from the snapshot that answered, not the repo cache dir by @danielhanchen in Read the GGUF context length from the snapshot that answered, not the repo cache dir #7920
• Tighten remaining artifact uploads (follow-up to Tauri smoke: stop uploading the whole target/debug tree (350GB of artifacts) #7930) by @danielhanchen in Tighten remaining artifact uploads (follow-up to #7930) #7939
• Studio: prevent long streaming Markdown stalls by @oobabooga in Studio: prevent long streaming Markdown stalls #7892
• Fix studio server crash when launched without a console (Windows) by @legobele in Fix studio server crash when launched without a console (Windows) #7932
• Fix GGUF export writing outside the chosen save directory on Windows ([Bug] Unsloth Studio Desktop Stuck @ 100% when training completes. #7897) by @danielhanchen in Fix GGUF export writing outside the chosen save directory on Windows (#7897) #7937
• Fix Windows PowerShell resource path authorization by @Imagineer99 in Fix Windows PowerShell resource path authorization #7943
• Desktop: fix the mobile-layout fallout of Windows text scaling by @danielhanchen in Desktop: fix the mobile-layout fallout of Windows text scaling #7927
• Desktop: make the close button quit outside macOS by @danielhanchen in Desktop: make the close button quit outside macOS #7929
• Desktop: pin the shipped backend package version so a shared venv gets upgraded by @danielhanchen in Desktop: pin the shipped backend package version so a shared venv gets upgraded #7926
• Studio: screen the detached-window test against the Windows blocklist too by @danielhanchen in Studio: screen the detached-window test against the Windows blocklist too #7936
• Require every declared denoiser, and every shard of it, before a cached pipeline reads as complete by @danielhanchen in Require every declared denoiser, and every shard of it, before a cached pipeline reads as complete #7871
• Fix the Windows-only snapshot path assertion in the diffusion prefetch test by @danielhanchen in Fix the Windows-only snapshot path assertion in the diffusion prefetch test #7933
• Studio: paint the image and video pickers from the curated catalog by @danielhanchen in Studio: paint the image and video pickers from the curated catalog #7896
• Bump install.sh / install.ps1 pins to unsloth>=2026.8.5, unsloth-zoo>=2026.8.4 by @danielhanchen in Bump install.sh / install.ps1 pins to unsloth>=2026.8.5, unsloth-zoo>=2026.8.4 #7967
• Studio: consistent rounded hover boxes for nav icon buttons by @shimmyshimmer in Studio: consistent rounded hover boxes for nav icon buttons #7970
• Studio: lead the Context Length note with the automatic fit by @shimmyshimmer in Studio: lead the Context Length note with the automatic fit #7988
• Studio: test the Windows launcher shape, not just the path normalizer by @danielhanchen in Studio: test the Windows launcher shape, not just the path normalizer #7969
• Studio: stop model rows staying lit after the dots menu closes by @shimmyshimmer in Studio: stop model rows staying lit after the dots menu closes #7987
• Update the nav icon button size contract to the 30px box by @danielhanchen in Update the nav icon button size contract to the 30px box #7979
• Studio: fetch the gated diffusion bases from ungated unsloth mirrors by @danielhanchen in Studio: fetch the gated diffusion bases from ungated unsloth mirrors #7952
• Studio: fix training stuck at 100% when the worker will not exit by @danielhanchen in Studio: fix training stuck at 100% when the worker will not exit #7938
• Report the real cause when a save or GGUF export fails by @danielhanchen in Report the real cause when a save or GGUF export fails #7861
• Locate the dictation model search box by test id, not by its copy by @danielhanchen in Locate the dictation model search box by test id, not by its copy #7914
• CI: drop three macOS legs that assert nothing a Linux runner does not by @danielhanchen in CI: drop three macOS legs that assert nothing a Linux runner does not #7974
• CI: fold the mac API smoke into the UI smoke, drop the subsumed install leg by @danielhanchen in CI: fold the mac API smoke into the UI smoke, drop the subsumed install leg #7975
• Installer: preserve rollback state and merge on partial venv move ([Bug] install.ps1: a failed venv rollback move on Windows splits the environment in two and strands it #7810) by @namdamdoi68-oss in Installer: preserve rollback state and merge on partial venv move (#7810) #7891
• CI: fold the mac update smoke into the same job, running last as teardown by @danielhanchen in CI: fold the mac update smoke into the same job, running last as teardown #7980
• CI: save caches only on main, so PR runs stop evicting the caches they need by @danielhanchen in CI: save caches only on main, so PR runs stop evicting the caches they need #7990
• Studio: drop the streaming animation wrappers once a message completes by @danielhanchen in Studio: drop the streaming animation wrappers once a message completes #7949
• Skip a test module with a missing optional dep instead of exiting pytest by @danielhanchen in Skip a test module with a missing optional dep instead of exiting pytest #7971
• Studio: restore the MLX gradient-norm chart by matching CUDA's clipping default by @Lyxot in Studio: restore the MLX gradient-norm chart by matching CUDA's clipping default #7917
• Studio: stop a GGUF pick from pulling the wrong companion bytes by @danielhanchen in Studio: stop a GGUF pick from pulling the wrong companion bytes #7870
• Studio: rework the media pane headings and tidy the Create controls by @shimmyshimmer in Studio: rework the media pane headings and tidy the Create controls #7986
• Studio: keep the sidebar fade in sync on resize and centre the profile row by @shimmyshimmer in Studio: keep the sidebar fade in sync on resize and centre the profile row #7977
• CI: run the three mac inference smokes as phases of one job by @danielhanchen in CI: run the three mac inference smokes as phases of one job #7978
• Studio: link the Train page across to image training by @shimmyshimmer in Studio: link the Train page across to image training #7996
• fix(studio/install): keep the rollback reference when a venv move stops partway ([Bug] install.ps1: a failed venv rollback move on Windows splits the environment in two and strands it #7810) by @Anai-Guo in fix(studio/install): keep the rollback reference when a venv move stops partway (#7810) #7823
• Studio: format profile stat numbers with the chosen locale by @danielhanchen in Studio: format profile stat numbers with the chosen locale #7916
• Settle deferred compile-mode switches between training steps by @danielhanchen in Settle deferred compile-mode switches between training steps #7955
• Studio: simplify desktop setup progress UI by @wasimysaid in Studio: simplify desktop setup progress UI #8011
• Tighten the gated-base preflight comments by @danielhanchen in Tighten the gated-base preflight comments #8003
• CI: point rust-cache at the real v2.9.1 tag and label the unlabelled pins by @danielhanchen in CI: point rust-cache at the real v2.9.1 tag and label the unlabelled pins #8005
• CI: move macOS runners off the retiring macos-14 image by @danielhanchen in CI: move macOS runners off the retiring macos-14 image #8004
• Studio: fetch the sd.cpp companion assets from unsloth mirrors, not community repacks by @danielhanchen in Studio: fetch the sd.cpp companion assets from unsloth mirrors, not community repacks #7995
• Studio: make the fp8 smoke probe catch a missing activation scale floor by @danielhanchen in Studio: make the fp8 smoke probe catch a missing activation scale floor #8008
• Diffusion: four fixes salvaged from the stale Diffusion: auto TE/VAE precision, video DiT quant correctness, and HunyuanVideo-1.5 attention speedup #7021 by @danielhanchen in Diffusion: four fixes salvaged from the stale #7021 #8006
• studio: persist the settings dialog picks that reset on every tab switch by @mahiatlinux in studio: persist the settings dialog picks that reset on every tab switch #7991
• Studio: bound every teardown await so a swallowed cancel cannot wedge the API by @mahiatlinux in Studio: bound every teardown await so a swallowed cancel cannot wedge the API #7881
• feat(studio): rework train page setup flow by @Sneakr in feat(studio): rework train page setup flow #7633
• Give the Windows sandbox bash a userland, and honour TEMP/TMP by @danielhanchen in Give the Windows sandbox bash a userland, and honour TEMP/TMP #8009
• Screen the cmd spellings the blocklist lexer quotes by @danielhanchen in Screen the cmd spellings the blocklist lexer quotes #8010
• CI: extract the POSIX local install into a composite action by @danielhanchen in CI: extract the POSIX local install into a composite action #8015
• Add option to launch Studio without opening the default browser by @shimmyshimmer in Add option to launch Studio without opening the default browser #7016
• Stop the rapid-submit chat UI check failing on a fast runner by @danielhanchen in Stop the rapid-submit chat UI check failing on a fast runner #7972
• CI: run the three Windows inference smokes as phases of one job by @danielhanchen in CI: run the three Windows inference smokes as phases of one job #8013
• fix(studio/rocm): don't install for a shadowing iGPU on mixed AMD hosts ([Bug] AMD GPU not loading into VRAM or detected #7776) by @Anai-Guo in fix(studio/rocm): don't install for a shadowing iGPU on mixed AMD hosts (#7776) #7778
• Diffusion: replace the attention perf claims with re-measured numbers by @danielhanchen in Diffusion: replace the attention perf claims with re-measured numbers #8021
• Desktop: fix Linux AppImage compatibility on newer distributions by @oobabooga in Desktop: fix Linux AppImage compatibility on newer distributions #7953
• Model hub: report why Hugging Face is unreachable instead of claiming you are offline by @danielhanchen in Model hub: report why Hugging Face is unreachable instead of claiming you are offline #7829
• Studio: stop two backend tests from asking the host what the answer is by @danielhanchen in Studio: stop two backend tests from asking the host what the answer is #8022
• Studio: do not let the default tool policy reject every plain Anthropic message by @danielhanchen in Studio: do not let the default tool policy reject every plain Anthropic message #8023
• Fix desktop window chrome by @wasimysaid in Fix desktop window chrome #8025
• Order the Mac inference phases so the gates stop fighting each other by @danielhanchen in Order the Mac inference phases so the gates stop fighting each other #8024
• Extract the API-only Studio boot into one script by @danielhanchen in Extract the API-only Studio boot into one script #8029
• Fix the desktop clean-machine CI: asset patterns and the retired Installer: stop requiring a developer toolchain on the consumer path #7547 pins by @danielhanchen in Fix the desktop clean-machine CI: asset patterns and the retired #7547 pins #8026
• Studio: center sheet header close buttons by @oobabooga in Studio: center sheet header close buttons #7957
• feat(studio): add settings-managed remote access by @Lyxot in feat(studio): add settings-managed remote access #7875
• Studio: add opt-in DSpark speculative decoding by @oobabooga in Studio: add opt-in DSpark speculative decoding #7968
• Bump install.sh / install.ps1 pin to unsloth>=2026.8.6, unsloth-zoo>=2026.8.5 by @danielhanchen in Bump install.sh / install.ps1 pin to unsloth>=2026.8.6, unsloth-zoo>=2026.8.5 #8038
• fix studio responsiveness across desktop and mobile web by @mahiatlinux in fix studio responsiveness across desktop and mobile web #8016
• Revert "Add option to launch Studio without opening the default browser" by @danielhanchen in Revert "Add option to launch Studio without opening the default browser" #8040
• Bump install.sh / install.ps1 pin to unsloth>=2026.8.7 by @danielhanchen in Bump install.sh / install.ps1 pin to unsloth>=2026.8.7 #8041
• Studio: send the GGUF filename when a collapsed On Device row is picked by @oobabooga in Studio: send the GGUF filename when a collapsed On Device row is picked #8053
• Validate an explicit GGUF variant before unloading the resident model by @NilayYadav in Validate an explicit GGUF variant before unloading the resident model #7862
• fix(studio): stop showing manifest-less cache entries as downloading ([Bug] Training overlay shows cached resources as Downloading at 99% without an active download #7858) by @Anai-Guo in fix(studio): stop showing manifest-less cache entries as downloading (#7858) #7864
• release-desktop: skip the Defender scan when Defender is unavailable by @danielhanchen in release-desktop: skip the Defender scan when Defender is unavailable #8059
• Report min_length as 0 when no sample has content by @vineethsaivs in Report min_length as 0 when no sample has content #7872
• Stop the health watchdog bypassing the backend startup grace period by @NilayYadav in Stop the health watchdog bypassing the backend startup grace period #7958
• Studio: honor MLX KV cache quantization and chat template overrides by @Lyxot in Studio: honor MLX KV cache quantization and chat template overrides #8007
• test: isolate the no-version gfx-override test from a real /opt/rocm by @danielhanchen in test: isolate the no-version gfx-override test from a real /opt/rocm #7397
• Studio: stop /api/inference/status raising on every loaded GGUF by @danielhanchen in Studio: stop /api/inference/status raising on every loaded GGUF #8074
• CI: share the gemma-3-270m-it model cache across Linux, macOS and Windows by @danielhanchen in CI: share the gemma-3-270m-it model cache across Linux, macOS and Windows #8068
• CI: share the Qwen3.5-2B tool-calling GGUF cache across Linux, macOS and Windows by @danielhanchen in CI: share the Qwen3.5-2B tool-calling GGUF cache across Linux, macOS and Windows #8069
• Fix the four Studio reds on main by @danielhanchen in Fix the four Studio reds on main #8072
• Studio: harden DSpark/MTP drafter detection and companion reclaim by @danielhanchen in Studio: harden DSpark/MTP drafter detection and companion reclaim #8037
• Studio: keep the stdio MCP session alive when a tool call is cancelled by @NilayYadav in Studio: keep the stdio MCP session alive when a tool call is cancelled #7625
• Reuse the torch2.10 prebuilt accelerator wheels on torch 2.12 by @danielhanchen in Reuse the torch2.10 prebuilt accelerator wheels on torch 2.12 #7495
• Studio: line up every sidebar row pill with New Chat by @shimmyshimmer in Studio: line up every sidebar row pill with New Chat #8079
• Normalize the /api/health wait and extract it into one script by @danielhanchen in Normalize the /api/health wait and extract it into one script #8060
• Make a fatal X11 error visible in the Linux desktop legs by @danielhanchen in Make a fatal X11 error visible in the Linux desktop legs #8063
• Studio: call XInitThreads so the Linux desktop app stops exiting on startup by @danielhanchen in Studio: call XInitThreads so the Linux desktop app stops exiting on startup #8075
• Studio: bound Deep Research first output waits by @alkinun in Studio: bound Deep Research first output waits #7998
• Allow current MLX dependencies with valid Studio sidecars by @Lyxot in Allow current MLX dependencies with valid Studio sidecars #7061
• Desktop: keep slide-in panels below the custom titlebar by @oobabooga in Desktop: keep slide-in panels below the custom titlebar #7954
• CI: stop running workflows on the pip branch, and mirror push path filters for the non-Studio workflows by @danielhanchen in CI: stop running workflows on the pip branch, and mirror push path filters for the non-Studio workflows #8085
• CI: always run Studio export/load-orchestrator and the security audit by @danielhanchen in CI: always run Studio export/load-orchestrator and the security audit #8086
• release-desktop: cache trusted-signing-cli and add job timeouts by @danielhanchen in release-desktop: cache trusted-signing-cli and add job timeouts #8087
• Show the public model id in the selector when the loaded model isn't in the picker list by @LeoBorcherding in Show the public model id in the selector when the loaded model isn't in the picker list #7966
• Explain why the gated-deltanet fast kernels were disabled by @danielhanchen in Explain why the gated-deltanet fast kernels were disabled #8073
• Studio: fix the Windows desktop setup log mojibake and double-printed steps by @danielhanchen in Studio: fix the Windows desktop setup log mojibake and double-printed steps #8083
• Installer: keep status messages off the progress bar line by @oobabooga in Installer: keep status messages off the progress bar line #8052
• Fix the status runtime-fields contract test left stale by Studio: stop /api/inference/status raising on every loaded GGUF #8074 by @shimmyshimmer in Fix the status runtime-fields contract test left stale by #8074 #8088
• Studio: pin the /status chat template echo behaviour by @shimmyshimmer in Studio: pin the /status chat template echo behaviour #8080
• Desktop: Fix unresponsive model selector on Images/Video pages on macOS by @NilayYadav in Desktop: Fix unresponsive model selector on Images/Video pages on macOS #8048
• Fix Swagger UI and ReDoc being blanked by the studio CSP by @NilayYadav in Fix Swagger UI and ReDoc being blanked by the studio CSP #8042
• Stop referencing a TAURI signing password secret that does not exist by @danielhanchen in Stop referencing a TAURI signing password secret that does not exist #8095
• CI: add an inert cross-repo capacity sweeper (do not merge yet) by @danielhanchen in CI: add an inert cross-repo capacity sweeper (do not merge yet) #8071
• tests: extend the version-compat matrices to the current PyPI releases by @danielhanchen in tests: extend the version-compat matrices to the current PyPI releases #8020
• Scope the security suite's offline guard to the tests that want it by @danielhanchen in Scope the security suite's offline guard to the tests that want it #8054
• Compare the drafter path with forward slashes on Windows too by @danielhanchen in Compare the drafter path with forward slashes on Windows too #8056
• Desktop: start at login, hidden in the menu bar by @NilayYadav in Desktop: start at login, hidden in the menu bar #8033
• CI: stop main and nightly runs cancelling each other by @danielhanchen in CI: stop main and nightly runs cancelling each other #8070
• Studio: stop the macOS launcher killing a warming backend, and the Mac tab blackout by @danielhanchen in Studio: stop the macOS launcher killing a warming backend, and the Mac tab blackout #8076
• Security: baseline seven reviewed scan_packages findings by @danielhanchen in Security: baseline seven reviewed scan_packages findings #8096
• Studio: label the chat model bar by repo, not the HF cache path by @danielhanchen in Studio: label the chat model bar by repo, not the HF cache path #8090
• Security: pin the credential-send allowlist entry to the reviewed file by @danielhanchen in Security: pin the credential-send allowlist entry to the reviewed file #8104
• CI: move the capacity sweeper off the retiring ubuntu-22.04 image by @danielhanchen in CI: move the capacity sweeper off the retiring ubuntu-22.04 image #8105
• release-desktop: add a VirusTotal pre-flight scan of the release bundles by @danielhanchen in release-desktop: add a VirusTotal pre-flight scan of the release bundles #8089
• release-desktop: fail closed when trusted-signing-cli is missing or broken by @danielhanchen in release-desktop: fail closed when trusted-signing-cli is missing or broken #8106
• Desktop: avoid WebKitGTK dmabuf crashes on Wayland by @wasimysaid in Desktop: avoid WebKitGTK dmabuf crashes on Wayland #8028
• Studio: cap llama.cpp source build parallelism by RAM by @shimmyshimmer in Studio: cap llama.cpp source build parallelism by RAM #8061
• Match the LM Studio GGUF suffix case insensitively by @danielhanchen in Match the LM Studio GGUF suffix case insensitively #8100
• Bump the actions group across 1 directory with 11 updates by @dependabot[bot] in Bump the actions group across 1 directory with 11 updates #8101
• Guard Windows Studio installs against active runtimes by @Etherll in Guard Windows Studio installs against active runtimes #7764
• Desktop: prevent thin AppImage environment failures by @oobabooga in Desktop: prevent thin AppImage environment failures #8055
• Studio: preserve the Windows launcher during updates by @Etherll in Studio: preserve the Windows launcher during updates #8092
• Desktop: fit windows within the monitor work area by @oobabooga in Desktop: fit windows within the monitor work area #8047
• Studio: keep dictation status reads off the sidecar lock by @shimmyshimmer in Studio: keep dictation status reads off the sidecar lock #8017
• Desktop: Fix "Copy path" failing in the desktop app by @NilayYadav in Desktop: Fix "Copy path" failing in the desktop app #7959
• Uninstall: remove WebView runtime data so reinstalls don't serve a stale frontend by @shimmyshimmer in Uninstall: remove WebView runtime data so reinstalls don't serve a stale frontend #7360
• Fail CI on new yes/no prompts in the installers by @danielhanchen in Fail CI on new yes/no prompts in the installers #8094
• Studio: continue a response that stopped early by @shimmyshimmer in Studio: continue a response that stopped early #8064
• Studio: say what a failed launcher move-aside costs by @danielhanchen in Studio: say what a failed launcher move-aside costs #8109
• Keep up with transformers 5.x, TRL 0.22.x and peft by @danielhanchen in Keep up with transformers 5.x, TRL 0.22.x and peft #7866
• Setup: clear stale WebView caches on install and update, keep user settings and data by @shimmyshimmer in Setup: clear stale WebView caches on install and update, keep user settings and data #7361
• Studio: find models already on the device before downloading, and make the first download cancellable by @shimmyshimmer in Studio: find models already on the device before downloading, and make the first download cancellable #7379
• [Bug] Unsloth Studio: let fenced HTML canvases honor the canvas network-access setting by @LeoBorcherding in [Bug] Unsloth Studio: let fenced HTML canvases honor the canvas network-access setting #8098
• Studio: keep chat titles whole so a wider sidebar shows more of them by @shimmyshimmer in Studio: keep chat titles whole so a wider sidebar shows more of them #8078
• Studio: keep the composer tool pills on one line by @shimmyshimmer in Studio: keep the composer tool pills on one line #8111
• Preflight the Codex GGUF check before loading the model by @NilayYadav in Preflight the Codex GGUF check before loading the model #7873
• Studio: keep the upgrade intent when the pip fallback runs by @danielhanchen in Studio: keep the upgrade intent when the pip fallback runs #8112
• Improve desktop installer progress UX by @wasimysaid in Improve desktop installer progress UX #8102
• studio: expose llama-server batch and micro-batch sizes as first-class load settings by @mahiatlinux in studio: expose llama-server batch and micro-batch sizes as first-class load settings #7973
• Bump install.sh / install.ps1 pin to unsloth>=2026.8.8 by @danielhanchen in Bump install.sh / install.ps1 pin to unsloth>=2026.8.8 #8114
• Clamp Deep Research max_tokens to loaded context window by @Souravrajvi0 in Clamp Deep Research max_tokens to loaded context window #7985
• Desktop: block HTTP loopback requests from untrusted images by @oobabooga in Desktop: block HTTP loopback requests from untrusted images #8046
• Studio: keep inference status probes off the event loop by @oobabooga in Studio: keep inference status probes off the event loop #7944
• Add Model memory settings to keep a loaded model in VRAM by @shimmyshimmer in Add Model memory settings to keep a loaded model in VRAM #8002
• perf(studio): make local model inventory scans linear by @Lyxot in perf(studio): make local model inventory scans linear #8043
• Repair seven tests left behind by their own subjects by @danielhanchen in Repair seven tests left behind by their own subjects #8132
• fix(security): re-baseline the huggingface-hub HTTP backoff loop for 1.x by @danielhanchen in fix(security): re-baseline the huggingface-hub HTTP backoff loop for 1.x #8135
• Studio: check llama.cpp cache access before setup by @oobabooga in Studio: check llama.cpp cache access before setup #8032
• studio: fix deep research progress reporting, loopback auth and model retries by @mahiatlinux in studio: fix deep research progress reporting, loopback auth and model retries #8129
• Reserve host RAM for AMD unified-memory APUs on the ROCm probe by @danielhanchen in Reserve host RAM for AMD unified-memory APUs on the ROCm probe #8131
• fix(AMD): let diffusion load on Windows ROCm, and say what a gated repo needs by @LeoBorcherding in fix(AMD): let diffusion load on Windows ROCm, and say what a gated repo needs #7981
• Desktop: ensure backend ownership for Tauri installs by @oobabooga in Desktop: ensure backend ownership for Tauri installs #8034
• Studio: keep row actions clear of overlay scrollbars by @oobabooga in Studio: keep row actions clear of overlay scrollbars #8031
• Report a denied UNSLOTH_LOCAL_LLAMA_CPP_DIR instead of aborting on errexit by @danielhanchen in Report a denied UNSLOTH_LOCAL_LLAMA_CPP_DIR instead of aborting on errexit #8143
• Studio: per-request perf stats and slot usage in the API monitor by @NilayYadav in Studio: per-request perf stats and slot usage in the API monitor #8045
• fix(studio): harden training setup, lifecycle, and audio loading by @Sneakr in fix(studio): harden training setup, lifecycle, and audio loading #8103
• Pin the accelerator gates in the DiT precision preflight test by @danielhanchen in Pin the accelerator gates in the DiT precision preflight test #8155
• Studio: fix STT download progress and cache safety by @oobabooga in Studio: fix STT download progress and cache safety #8035
• Fix desktop image drops for chat attachments by @Souravrajvi0 in Fix desktop image drops for chat attachments #7982
• Studio: fall back to CPU when auto-selected Vulkan crashes by @oobabooga in Studio: fall back to CPU when auto-selected Vulkan crashes #8050
• Studio: stop research threads 409-ing on autosave, keep server-managed messages intact by @NilayYadav in Studio: stop research threads 409-ing on autosave, keep server-managed messages intact #7956
• Studio: retry Xet once on a stall before falling back to HTTP by @danielhanchen in Studio: retry Xet once on a stall before falling back to HTTP #8166
• Deduplicate the scan-packages baseline, and gate it with tests by @danielhanchen in Deduplicate the scan-packages baseline, and gate it with tests #8110
• Add Markdown conversation export by @ousamabenyounes in Add Markdown conversation export #7664
• Bump the cargo-tauri group across 1 directory with 18 updates by @dependabot[bot] in Bump the cargo-tauri group across 1 directory with 18 updates #8107
• CI: prune superseded Actions caches by @danielhanchen in CI: prune superseded Actions caches #8167
• Publish studio_install_id without clobbering a concurrent writer by @danielhanchen in Publish studio_install_id without clobbering a concurrent writer #8145
• Desktop: refuse to republish an existing release version by @oobabooga in Desktop: refuse to republish an existing release version #7941
• Studio: surface API errors and align audio limits by @oobabooga in Studio: surface API errors and align audio limits #7946
• Stop the CUDA test spoofs reporting an exhausted card by @danielhanchen in Stop the CUDA test spoofs reporting an exhausted card #8123
• Keep download scopes out of the GGUF quantization list by @shimmyshimmer in Keep download scopes out of the GGUF quantization list #8153
• Do not choose flex attention on a card that cannot run its kernel by @danielhanchen in Do not choose flex attention on a card that cannot run its kernel #8137
• Render a static prefix when default_system_message is None by @vineethsaivs in Render a static prefix when default_system_message is None #8117
• Studio: rate the API monitor Throughput tile on the decode window by @shimmyshimmer in Studio: rate the API monitor Throughput tile on the decode window #8139
• [Fix] Unsloth Studio: Strip ANSI escape codes from Studio tool output panes by @Souravrajvi0 in [Fix] Unsloth Studio: Strip ANSI escape codes from Studio tool output panes #7976
• Attach save_lora without a vLLM engine by @danielhanchen in Attach save_lora without a vLLM engine #8127
• Studio: bring back the sidebar scrollbar without losing row width by @shimmyshimmer in Studio: bring back the sidebar scrollbar without losing row width #8147
• Studio: exempt shared top-level trees from the integrity check for our own wheels too by @danielhanchen in Studio: exempt shared top-level trees from the integrity check for our own wheels too #8179
• Do not stage diffusion files that are already in the cache by @shimmyshimmer in Do not stage diffusion files that are already in the cache #8154
• Catch the llama-prebuilt validator stubs up with rocm_gfx by @danielhanchen in Catch the llama-prebuilt validator stubs up with rocm_gfx #8182
• Pin the auto-load success toast by its branch, not by one spelling by @danielhanchen in Pin the auto-load success toast by its branch, not by one spelling #8183
• Studio: route every Windows installer line through the UTF-8 stdout sink by @danielhanchen in Studio: route every Windows installer line through the UTF-8 stdout sink #8148
• Studio: hold the Mac capability verdict while an MLX repair can still overturn it by @danielhanchen in Studio: hold the Mac capability verdict while an MLX repair can still overturn it #8152
• Count a RECORD path claim before the row is filtered out by @Imagineer99 in Count a RECORD path claim before the row is filtered out #8181
• Fix on-device GGUF image/video picks asking for a quantization by @shimmyshimmer in Fix on-device GGUF image/video picks asking for a quantization #8149
• Enforce the interface's video resolutions in the API too by @danielhanchen in Enforce the interface's video resolutions in the API too #8159
• Studio: add a loaded models indicator with per-model eject by @shimmyshimmer in Studio: add a loaded models indicator with per-model eject #8082
• Desktop: show a closing overlay while the backend shuts down by @danielhanchen in Desktop: show a closing overlay while the backend shuts down #8151
• Studio: make the gallery video link absolute so the desktop app can play and save it by @danielhanchen in Studio: make the gallery video link absolute so the desktop app can play and save it #8173
• Bump install.sh / install.ps1 pin to unsloth>=2026.8.9 by @danielhanchen in Bump install.sh / install.ps1 pin to unsloth>=2026.8.9 #8189
• Studio: eliminate Cargo and Clippy warnings by @wasimysaid in Studio: eliminate Cargo and Clippy warnings #8192
• studio: serve the deep research event stream over post so proxies stop buffering it by @mahiatlinux in studio: serve the deep research event stream over post so proxies stop buffering it #8205
• Desktop release: build the three platform legs in parallel by @danielhanchen in Desktop release: build the three platform legs in parallel #8191
• Desktop release: start the publish job alongside the build matrix by @danielhanchen in Desktop release: start the publish job alongside the build matrix #8193
• tests: make pytest tests collect cleanly and drop the deleted QVQ registry mirror by @danielhanchen in tests: make pytest tests collect cleanly and drop the deleted QVQ registry mirror #8206
• Desktop release: run the VirusTotal scan after publishing, in its own job by @danielhanchen in Desktop release: run the VirusTotal scan after publishing, in its own job #8194
• tests: build the hub-error fixtures without calling init (huggingface_hub 1.x) by @danielhanchen in tests: build the hub-error fixtures without calling __init__ (huggingface_hub 1.x) #8207
• Studio: stop MTP forcing llama-server to a single parallel slot by @danielhanchen in Studio: stop MTP forcing llama-server to a single parallel slot #8172
• test(studio): stop the backend suite depending on a live Hub by @danielhanchen in test(studio): stop the backend suite depending on a live Hub #8174
• Repin six UI contracts on what they guard, not how it was spelled by @danielhanchen in Repin six UI contracts on what they guard, not how it was spelled #8197
• Studio: only dodge an overlay box that reaches the corner stack by @shimmyshimmer in Studio: only dodge an overlay box that reaches the corner stack #8210
• Studio: surface an MLX completion-masking miss as a warning, not a status line by @BardiaKoopah in Studio: surface an MLX completion-masking miss as a warning, not a status line #8133
• Desktop: block accidental reloads in the Windows WebView by @oobabooga in Desktop: block accidental reloads in the Windows WebView #7945
• Studio: preserve training config values across YAML load and save by @oobabooga in Studio: preserve training config values across YAML load and save #7942
• Stop a finished download reporting 0 B and offering Retry by @danielhanchen in Stop a finished download reporting 0 B and offering Retry #8160
• Install a CUDA-matched xFormers on Windows instead of whatever PyPI serves by @danielhanchen in Install a CUDA-matched xFormers on Windows instead of whatever PyPI serves #8156
• fix(studio): release the inference subprocess's VRAM when its last model unloads by @LeoBorcherding in fix(studio): release the inference subprocess's VRAM when its last model unloads #8221
• fix(studio): reserve a bounded OS headroom on unified ROCm APUs, not a flat 20% by @LeoBorcherding in fix(studio): reserve a bounded OS headroom on unified ROCm APUs, not a flat 20% #7960
• Studio: fix Windows desktop updates failing integrity checks by @Etherll in Studio: fix Windows desktop updates failing integrity checks #8185
• Report the precision actually in use, and refuse an explicit one we cannot honor by @danielhanchen in Report the precision actually in use, and refuse an explicit one we cannot honor #8165
• Pin the training precision, video preset and image metadata contracts by @danielhanchen in Pin the training precision, video preset and image metadata contracts #8162
• Say when xFormers is installed but its kernels cannot load by @danielhanchen in Say when xFormers is installed but its kernels cannot load #8157
• Revert "Say when xFormers is installed but its kernels cannot load" by @danielhanchen in Revert "Say when xFormers is installed but its kernels cannot load" #8237
• Studio: stream oversized diffusion components on low-VRAM GPUs by @oobabooga in Studio: stream oversized diffusion components on low-VRAM GPUs #8141
• fix(studio): wire llama's OpenMP runtime into slim whisper installs by @danielhanchen in fix(studio): wire llama's OpenMP runtime into slim whisper installs #8230
• fix(studio): verify TLS against the OS trust store at runtime (corporate TLS-inspection proxies) by @Cesarsk in fix(studio): verify TLS against the OS trust store at runtime (corporate TLS-inspection proxies) #8108
• Tell users to upgrade the dependency, not transformers, when a --no-deps install leaves a floor unmet by @danielhanchen in Tell users to upgrade the dependency, not transformers, when a --no-deps install leaves a floor unmet #8212
• Repo tests: pin the post-publish VirusTotal contract instead of the old pre-flight one by @danielhanchen in Repo tests: pin the post-publish VirusTotal contract instead of the old pre-flight one #8240
• Let the pipeline gate survive a diffusers whose lazy submodule cannot import by @danielhanchen in Let the pipeline gate survive a diffusers whose lazy submodule cannot import #8227
• Stop the idle-unload TTL test racing a loaded runner by @danielhanchen in Stop the idle-unload TTL test racing a loaded runner #8175
• Let the floating monitor paint over the notification stack by @danielhanchen in Let the floating monitor paint over the notification stack #8199
• Format the release permissions test to the kwarg-spacing hook by @danielhanchen in Format the release permissions test to the kwarg-spacing hook #8245
• Format the release-permissions test so pre-commit.ci stops failing on main by @danielhanchen in Format the release-permissions test so pre-commit.ci stops failing on main #8247
• add folder picking and drag-drop for paired image and caption datasets by @mahiatlinux in add folder picking and drag-drop for paired image and caption datasets #8216
• Backfill the conversion symbols peft imports and transformers 5 dropped by @danielhanchen in Backfill the conversion symbols peft imports and transformers 5 dropped #8051
• Studio: only clean up GGUF artifacts the export owns by @oobabooga in Studio: only clean up GGUF artifacts the export owns #7940
• Resume diffusion training from a stop-and-save checkpoint by @danielhanchen in Resume diffusion training from a stop-and-save checkpoint #8163
• Stop padding-free SFT from tripping the TRL >= 1.0.0 max_length guard by @danielhanchen in Stop padding-free SFT from tripping the TRL >= 1.0.0 max_length guard #7951
• Fix Studio loaded-model GGUF quant label by @Imagineer99 in Fix Studio loaded-model GGUF quant label #8238
• Do not scan vendored packages that refuse to import on our floor by @danielhanchen in Do not scan vendored packages that refuse to import on our floor #8252
• Desktop: confirm quit during training by @NilayYadav in Desktop: confirm quit during training  #7790
• Fetch the llama.cpp validation model only when the smoke test reads it by @danielhanchen in Fetch the llama.cpp validation model only when the smoke test reads it #8248
• Make install.ps1 work with the user's PowerShell profile loaded by @danielhanchen in Make install.ps1 work with the user's PowerShell profile loaded #8161
• Name a torchvision whose compiled ops do not match torch by @danielhanchen in Name a torchvision whose compiled ops do not match torch #8128
• Diffusion: source every image family's companions from unsloth mirrors by @danielhanchen in Diffusion: source every image family's companions from unsloth mirrors #8030
• studio: infer splits for metadata-free cached datasets by @Ricardo-M-L in studio: infer splits for metadata-free cached datasets #8144
• Reject an incompatible GGUF and base pair before the download, not after by @danielhanchen in Reject an incompatible GGUF and base pair before the download, not after #8158
• Make the desktop release contract tests fail when the contract breaks by @danielhanchen in Make the desktop release contract tests fail when the contract breaks #8228
• GRPO: wrap the model that owns the head when falling back to hidden states by @danielhanchen in GRPO: wrap the model that owns the head when falling back to hidden states #8231
• Diffusion: allow fp8 on Qwen-Image again, and split the training deny by @danielhanchen in Diffusion: allow fp8 on Qwen-Image again, and split the training deny #8018
• Report the network error when the offline retry finds nothing cached by @danielhanchen in Report the network error when the offline retry finds nothing cached #8121
• Keep an explicitly requested float32 model in float32 without bfloat16 by @danielhanchen in Keep an explicitly requested float32 model in float32 without bfloat16 #7867
• Do not let a broken TensorFlow install break the Unsloth import by @danielhanchen in Do not let a broken TensorFlow install break the Unsloth import #8124
• Carry the sliding window into SDPA, and treat a zero window as no window by @danielhanchen in Carry the sliding window into SDPA, and treat a zero window as no window #8253
• Resolve latest mlx-vlm and route its new processors by @Lyxot in Resolve latest mlx-vlm and route its new processors #8251
• Studio: align Images and settings layouts by @wasimysaid in Studio: align Images and settings layouts #8164
• Studio: realign the image precision contracts with the fail-closed refusal by @danielhanchen in Studio: realign the image precision contracts with the fail-closed refusal #8257
• Studio: name the quant a diffusion GGUF pick actually loaded, instead of the compute dtype by @danielhanchen in Studio: name the quant a diffusion GGUF pick actually loaded, instead of the compute dtype #8241
• Studio: budget a pre-cast fp8 text encoder at the size it actually loads by @danielhanchen in Studio: budget a pre-cast fp8 text encoder at the size it actually loads #8218
• Desktop: keep probing the announced backend port instead of discarding it by @danielhanchen in Desktop: keep probing the announced backend port instead of discarding it #8211
• Stop the CPU fake-train tests from spoofing torch for the whole session by @danielhanchen in Stop the CPU fake-train tests from spoofing torch for the whole session #8259
• Studio: grey the sidebar dot a finished chat leaves by @shimmyshimmer in Studio: grey the sidebar dot a finished chat leaves #8258
• Studio: stop leaking child processes on an abnormal exit by @danielhanchen in Studio: stop leaking child processes on an abnormal exit #8170
• Restore SFT loss_type='nll' default over TRL 1.7.0 chunked_nll by @danielhanchen in Restore SFT loss_type='nll' default over TRL 1.7.0 chunked_nll #8261
• Assert the image family's pipeline class in the training preflight, before the GPU teardown by @danielhanchen in Assert the image family's pipeline class in the training preflight, before the GPU teardown #8236
• Studio: let image generation be stopped by @danielhanchen in Studio: let image generation be stopped #8219
• Studio: make the image memory plan resolution-aware and stop reserving the text encoder during sampling by @danielhanchen in Studio: make the image memory plan resolution-aware and stop reserving the text encoder during sampling #8224
• remove the unused images train dialog and fix media page responsiveness by @mahiatlinux in remove the unused images train dialog and fix media page responsiveness #8215
• Studio: size FLUX.2-klein 9B LoRA training from its own weights, not the 4B family number by @danielhanchen in Studio: size FLUX.2-klein 9B LoRA training from its own weights, not the 4B family number #8263
• Studio: LTX-2 video LoRA training from still images by @danielhanchen in Studio: LTX-2 video LoRA training from still images #8196
• Say when attention has no fused kernel, and log the request behind a failed video by @danielhanchen in Say when attention has no fused kernel, and log the request behind a failed video #8233
• Turn off embedding offload when the model is dispatched across devices by @danielhanchen in Turn off embedding offload when the model is dispatched across devices #8272
• Studio: keep the overlay stack off a maximised Live monitor by @danielhanchen in Studio: keep the overlay stack off a maximised Live monitor #8190
• Make the diffusion checkpoint-resume tests describe the host they run on by @danielhanchen in Make the diffusion checkpoint-resume tests describe the host they run on #8273
• Install the Linux CUDA sd-cli, and bump the pin two releases forward by @danielhanchen in Install the Linux CUDA sd-cli, and bump the pin two releases forward #8208
• Studio: add MiniMax H3 video generation by @oobabooga in Studio: add MiniMax H3 video generation #7989
• Bump install.sh / install.ps1 pin to unsloth>=2026.8.10 by @danielhanchen in Bump install.sh / install.ps1 pin to unsloth>=2026.8.10 #8276
• Stop a GGUF pick downloading the dense base transformer it never opens by @danielhanchen in Stop a GGUF pick downloading the dense base transformer it never opens #8232
• Studio: let a model load be cancelled while it is still loading by @danielhanchen in Studio: let a model load be cancelled while it is still loading #8243
• Studio: stop a multi-checkpoint GGUF repo collapsing into one row per quant by @danielhanchen in Studio: stop a multi-checkpoint GGUF repo collapsing into one row per quant #8222
• Studio: refuse a diffusion or video load that cannot fit unified memory by @danielhanchen in Studio: refuse a diffusion or video load that cannot fit unified memory #8213
• Studio: run MiniMax-H3's Qwen3-VL conditioner from the hosted INT8 checkpoint by @danielhanchen in Studio: run MiniMax-H3's Qwen3-VL conditioner from the hosted INT8 checkpoint #8283
• Fix image page pane and header alignment by @Imagineer99 in Fix image page pane and header alignment #8176
• Studio: match the loaded models card to the update banners by @shimmyshimmer in Studio: match the loaded models card to the update banners #8292
• Fix main: the two diffusers-import preflight tests only pass on a GPU host by @danielhanchen in Fix main: the two diffusers-import preflight tests only pass on a GPU host #8281
• Desktop: fix audio files being rejected when dropped on the chat by @NilayYadav in Desktop: fix audio files being rejected when dropped on the chat #8271
• Fix two macOS installer bugs by @NilayYadav in Fix two macOS installer bugs #8279
• Studio: offer a per-canvas network grant when the offline CSP blocks a preview by @NilayYadav in Studio: offer a per-canvas network grant when the offline CSP blocks a preview #8262
• Read PSModulePath under the key Windows actually exports by @danielhanchen in Read PSModulePath under the key Windows actually exports #8304
• Studio: stop the sidebar flashing an icon rail it never collapses to by @danielhanchen in Studio: stop the sidebar flashing an icon rail it never collapses to #8300
• Studio: keep the loaded models card in the corner it is supposed to be in by @danielhanchen in Studio: keep the loaded models card in the corner it is supposed to be in #8301
• Studio: say what the Train row is waiting for while it spins by @danielhanchen in Studio: say what the Train row is waiting for while it spins #8302
• Fix image model dependency downloads and show full footprint by @Imagineer99 in Fix image model dependency downloads and show full footprint #8036
• soften the dmg background glow and clear it off the icon label by @mahiatlinux in soften the dmg background glow and clear it off the icon label #8296
• Studio: remeasure the overlay stack when its content resizes in place by @danielhanchen in Studio: remeasure the overlay stack when its content resizes in place #8315
• Raise when SyntheticDataKit's vLLM server fails to start by @danielhanchen in Raise when SyntheticDataKit's vLLM server fails to start #8297
• Only trust the adjacent ROCm-on-WSL helper for a real checkout run by @danielhanchen in Only trust the adjacent ROCm-on-WSL helper for a real checkout run #8316
• Studio: confirm a web_search call that fetches a url in auto mode by @danielhanchen in Studio: confirm a web_search call that fetches a url in auto mode #8313
• Desktop: route native drops to the drop zone under the cursor by @NilayYadav in Desktop: route native drops to the drop zone under the cursor #8265
• Fix resume button missing after stop-and-save ([Bug] Resume training button missing immediately after Stop and Save in latest Unsloth Studio #8150) by @Souravrajvi0 in Fix resume button missing after stop-and-save (#8150) #8217
• Studio: MiniMax-H3 follow-ups from the 7989 review by @danielhanchen in Studio: MiniMax-H3 follow-ups from the 7989 review #8280
• Studio: run MiniMax-H3's INT8 denoiser from the ConvRot checkpoint by @danielhanchen in Studio: run MiniMax-H3's INT8 denoiser from the ConvRot checkpoint #8293
• Desktop: unify normal release updater flow by @wasimysaid in Desktop: unify normal release updater flow #8298
• Start the dmg halo taper under the icon so the disc still reads round by @danielhanchen in Start the dmg halo taper under the icon so the disc still reads round #8321
• Studio: show the files a tool call creates, and keep them in one place by @danielhanchen in Studio: show the files a tool call creates, and keep them in one place #8256
• Studio: keep a rehearsal tool call quoted in markdown code as text by @danielhanchen in Studio: keep a rehearsal tool call quoted in markdown code as text #8312
• CLI: verify-install reports a healthy install as broken around an editable checkout by @danielhanchen in CLI: verify-install reports a healthy install as broken around an editable checkout #8308
• Studio: a base quant published at several bit widths is several variants by @danielhanchen in Studio: a base quant published at several bit widths is several variants #8255
• Studio: make image-model companion assets reusable, visible, and safely removable by @danielhanchen in Studio: make image-model companion assets reusable, visible, and safely removable #8223
• Studio: let a dataset of clips be selected for training by @danielhanchen in Studio: let a dataset of clips be selected for training #8287
• studio: offer the undistilled z-image base for lora training by @mahiatlinux in studio: offer the undistilled z-image base for lora training #8291
• Studio: better default Images and Video prompts by @shimmyshimmer in Studio: better default Images and Video prompts #8324
• CI: avoid blocking Windows releases on cache saves by @wasimysaid in CI: avoid blocking Windows releases on cache saves #8325
• Studio: a binary request body should be a 422, not a 500 with the payload in the log by @danielhanchen in Studio: a binary request body should be a 422, not a 500 with the payload in the log #8307
• Studio: stop two lines repeating in the server log for the whole session by @danielhanchen in Studio: stop two lines repeating in the server log for the whole session #8306
• Studio: give the liveness polls one shared heartbeat instead of one each by @danielhanchen in Studio: give the liveness polls one shared heartbeat instead of one each #8310
• Studio: stop reporting the run's mean loss as the final step's loss by @danielhanchen in Studio: stop reporting the run's mean loss as the final step's loss #8314
• studio: render a chat send without waiting on persistence by @mahiatlinux in studio: render a chat send without waiting on persistence #8136
• Studio: enable video generation on Apple Silicon by @Lyxot in Studio: enable video generation on Apple Silicon #8198
• Studio: keep sidebar logo visible during training by @wasimysaid in Studio: keep sidebar logo visible during training #8331
• Absorb TRL config arguments retired between versions by @danielhanchen in Absorb TRL config arguments retired between versions #8330
• Studio: use Muse Glimmer's published sampling defaults by @danielhanchen in Studio: use Muse Glimmer's published sampling defaults #8332
• Studio: keep third-party progress bars and raw trainer prints out of the structured log by @danielhanchen in Studio: keep third-party progress bars and raw trainer prints out of the structured log #8311
• Video: make MiniMax-H3's Diffusers path fast by default by @danielhanchen in Video: make MiniMax-H3's Diffusers path fast by default #8320
• Studio: give the MiniMax-H3 reference denoiser an int8 and fp8 rung by @danielhanchen in Studio: give the MiniMax-H3 reference denoiser an int8 and fp8 rung #8329

New Contributors

• @namdamdoi68-oss made their first contribution in Installer: preserve rollback state and merge on partial venv move (#7810) #7891

Full Changelog: v0.1.52-beta...v0.1.60-beta

---

This discussion was created from the release Meta Muse Glimmer.