New UI Redesign + Qwen3.6 · unslothai unsloth · Discussion #5153 · GitHub

shimmyshimmer
Apr 23, 2026
Maintainer

Hey guys, we revamped the entire Unsloth Studio UI and UX experience to put an emphasis on chat and training:

Added a collapsible sidebar based on community feedback
You can now delete chats and search past conversations
New Preserve Thinking toggle for models that support it like Qwen3.6
Cleaner, more consistent design with easier navigation
Expanded Settings page with options to change your profile picture, name, and more
No more entering your Hugging Face token twice
gpt-oss now has low, medium and high thinking toggles.
Now uses latest llama.cpp prebuilt, even on Linux CUDA
Lots of bug, consistency and stability fixes
Kimi-K2.6 can now be run!
We also added experimental API support. Guides, announcement etc will come next week.

Qwen3.6 was also also previously already supported in Unsloth Studio for running and training. You can train and run Qwen3.6-27B right now!

What's Changed

Only run ldconfig CUDA-linking recovery when we have permission by @danielhanchen in Only run ldconfig CUDA-linking recovery when we have permission #4930
Fix Mistral DPO/preference training crash on non-xformers platforms (e.g. Intel XPU) by @cheehook in Fix Mistral DPO/preference training crash on non-xformers platforms (e.g. Intel XPU) #4889
Fix raw text paragraph break normalization by @kiankyars in Fix raw text paragraph break normalization #4884
Studio: keep chat input visible and fix compare pane clipping by @Imagineer99 in Studio: keep chat input visible and fix compare pane clipping #4924
fix: check find() return value before adding offset in try_fix_tokenizer by @Ricardo-M-L in fix: check find() return value before adding offset in try_fix_tokenizer #4923
updated models template mappers. added lfm2.5vl450m to transformers 5… by @rolandtannous in updated models template mappers. added lfm2.5vl450m to transformers 5… #4939
Revert "updated models template mappers. added lfm2.5vl450m to transformers 5…" by @rolandtannous in Revert "updated models template mappers. added lfm2.5vl450m to transformers 5…" #4945
Add AMD ROCm/HIP support across installer and hardware detection by @danielhanchen in Add AMD ROCm/HIP support across installer and hardware detection #4720
Pin bitsandbytes to continuous-release_main on ROCm (4-bit decode fix) by @danielhanchen in Pin bitsandbytes to continuous-release_main on ROCm (4-bit decode fix) #4954
Fix Gemma-4 GRPO catastrophic KL divergence with TRL 1.0.0+ by @danielhanchen in Fix Gemma-4 GRPO catastrophic KL divergence with TRL 1.0.0+ #4934
Add ROCm test suite (companion to Add AMD ROCm/HIP support across installer and hardware detection #4720) by @danielhanchen in Add ROCm test suite (companion to #4720) #4824
updating gemma4 script by @Manan17 in updating gemma4 script #4992
Move gemma4 script by @Manan17 in Move gemma4 script #4994
studio: fix route transition DOM duplication via AnimatePresence mode="wait" by @AdamPlatin123 in studio: fix route transition DOM duplication via AnimatePresence mode="wait" #4987
Studio: Prompt manager, message deletion, and chat UI improvements by @Imagineer99 in Studio: Prompt manager, message deletion, and chat UI improvements #4938
Pin kernels==0.12.1 to fix training import failure by @rolandtannous in Pin kernels==0.12.1 to fix training import failure #5000
Studio: Expose openai and anthropic compatible external API end points by @danielhanchen in Studio: Expose openai and anthropic compatible external API end points #4956
studio: skip training status/metrics polling when idle by @AdamPlatin123 in studio: skip training status/metrics polling when idle #4988
studio: fix api-keys access + refresh by @wasimysaid in studio: fix api-keys access + refresh #5005
Studio: Polish API key copy button and harden async clipboard fallback by @Imagineer99 in Studio: Polish API key copy button and harden async clipboard fallback #5006
fix(studio): default chart view to full training history by @Barath19 in fix(studio): default chart view to full training history #5007
[Studio] Show non exported models in chat UI by @Datta0 in [Studio] Show non exported models in chat UI #4892
[Studio] Install flash attn at setup time for linux by @Datta0 in [Studio] Install flash attn at setup time for linux #4979
fix(studio): remove 300s cap on load_checkpoint (inherits 3600s default) by @TF-MTGE in fix(studio): remove 300s cap on load_checkpoint (inherits 3600s default) #4922
Studio: honor explicit GGUF ctx and default to 4096 when weights exceed VRAM by @danielhanchen in Studio: honor explicit GGUF ctx and default to 4096 when weights exceed VRAM #5011
Studio: make GGUF disk-space preflight cache-aware by @danielhanchen in Studio: make GGUF disk-space preflight cache-aware #5012
Studio: anchor ctx-slider warning threshold at 4096 when weights exceed VRAM by @danielhanchen in Studio: anchor ctx-slider warning threshold at 4096 when weights exceed VRAM #5014
studio: show HF model download progress in training start overlay by @danielhanchen in studio: show HF model download progress in training start overlay #4894
studio: stream export worker output into the export dialog by @danielhanchen in studio: stream export worker output into the export dialog #4897
Fix num_items_in_batch GA for Gemma4 by @Datta0 in Fix num_items_in_batch GA for Gemma4 #4998
studio: pin peft to 0.18.1 to fix export subprocess issues by @rolandtannous in studio: pin peft to 0.18.1 to fix export subprocess issues #5015
Studio: live model-load progress + rate/ETA on download and load by @danielhanchen in Studio: live model-load progress + rate/ETA on download and load #5017
[pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in [pre-commit.ci] pre-commit autoupdate #5004
Fix bitsandbytes ROCm install by using pip instead of uv by @edamamez in Fix bitsandbytes ROCm install by using pip instead of uv #4966
Studio: split model-load progress label across two rows by @danielhanchen in Studio: split model-load progress label across two rows #5020
Studio: hard-stop at n_ctx with a 'Context limit reached' toast by @danielhanchen in Studio: hard-stop at n_ctx with a 'Context limit reached' toast #5021
[moe][gemma4] Target MoE for gemma4 by @Datta0 in [moe][gemma4] Target MoE for gemma4 #4913
Add configurable PyTorch mirror via UNSLOTH_PYTORCH_MIRROR env var by @rolandtannous in Add configurable PyTorch mirror via UNSLOTH_PYTORCH_MIRROR env var #5024
Studio: support GGUF variant selection for non-suffixed repos by @Imagineer99 in Studio: support GGUF variant selection for non-suffixed repos #5023
fix: prevent offline freeze by fixing stats retry and forwarding local_files_only by @DavidSolanas in fix: prevent offline freeze by fixing stats retry and forwarding local_files_only #5016
Respect classification head skip list on pre-quantized 4-bit checkpoints ([Bug] "normal_kernel_cuda" not implemented for 'Byte' for unsloth/Qwen3-4B-bnb-4bit #5027) by @danielhanchen in Respect classification head skip list on pre-quantized 4-bit checkpoints (#5027) #5034
fix(rocm): tighten gfx regex to ignore generic ISA lines by @danielhanchen in fix(rocm): tighten gfx regex to ignore generic ISA lines #5033
Fix grad-accum accepts_loss_kwargs detection for vision wrappers by @danielhanchen in Fix grad-accum accepts_loss_kwargs detection for vision wrappers #5036
grpo_compute_loss_slow called with wrong positional args by @jonahsamost in grpo_compute_loss_slow called with wrong positional args #4887
Gate trl disable_gradient_checkpointing warning on UNSLOTH_ENABLE_LOGGING by @danielhanchen in Gate trl disable_gradient_checkpointing warning on UNSLOTH_ENABLE_LOGGING #5038
Studio: refresh Downloaded GGUF list and recurse into variant subdirs by @danielhanchen in Studio: refresh Downloaded GGUF list and recurse into variant subdirs #5032
feat: Add support for OLMo-3 model by @OnePunchMonk in feat: Add support for OLMo-3 model #4678
feat: Add cactus QAT scheme support by @OnePunchMonk in feat: Add cactus QAT scheme support #4679
Re-apply updated models template mappers. added lfm2.5vl450m to transformers 5… #4939: updated models template mappers by @rolandtannous in Re-apply #4939: updated models template mappers #4950
Studio: add folder browser modal for Custom Folders by @danielhanchen in Studio: add folder browser modal for Custom Folders #5035
Bump Studio installer minimum to 2026.4.5 by @danielhanchen in Bump Studio installer minimum to 2026.4.5 #5041
fix Gemma4 flash attn disable by @mmathew23 in fix Gemma4 flash attn disable #5045
BUG: fix _fix_chat_template for ChatML templates missing add_generation_prompt ([Bug] Hermes Lora chat test bug: does not have a {% if add_generation_prompt %} for generation purposes #4150) by @kimimgo in BUG: fix _fix_chat_template for ChatML templates missing add_generation_prompt (#4150) #4426
fix: use direct registry API for PATH writes instead of SetEnvironmentVariable by @Etherll in fix: use direct registry API for PATH writes instead of SetEnvironmentVariable #4961
Chat-template repair: warn-by-default, AST classification, dict support by @danielhanchen in Chat-template repair: warn-by-default, AST classification, dict support #5049
Restrict flash attn to <=256 head dim. Consolidate attn impl checks by @Datta0 in Restrict flash attn to <=256 head dim. Consolidate attn impl checks #5051
Remove legacy venv Scripts entry from User PATH on upgrade by @danielhanchen in Remove legacy venv Scripts entry from User PATH on upgrade #5060
Fix review findings for chat-template repair (Chat-template repair: warn-by-default, AST classification, dict support #5049) by @danielhanchen in Fix review findings for chat-template repair (#5049) #5056
Studio: Ollama support, recommended folders, Custom Folders UX polish by @danielhanchen in Studio: Ollama support, recommended folders, Custom Folders UX polish #5050
feat(studio): replace navbar with collapsible sidebar by @wasimysaid in feat(studio): replace navbar with collapsible sidebar #4936
fix audio dataset preview and finetuning by @CodeMan62 in fix audio dataset preview and finetuning #5043
Chat first onboarding by @wasimysaid in Chat first onboarding #5063
Fix onboarding followups by @wasimysaid in Fix onboarding followups #5064
Studio: Default Gemma fallback for chat + AI assist by @Imagineer99 in Studio: Default Gemma fallback for chat + AI assist #5066
fix: multi-GPU inference crash for bnb 4-bit/8-bit models by @danielhanchen in fix: multi-GPU inference crash for bnb 4-bit/8-bit models #5068
Add Qwen3.6 inference defaults for Studio by @danielhanchen in Add Qwen3.6 inference defaults for Studio #5065
Add qwen3.6 script by @Manan17 in Add qwen3.6 script #5084
Studio: forward standard OpenAI tools / tool_choice to llama-server by @rolandtannous in Studio: forward standard OpenAI tools / tool_choice to llama-server #5099
fix(studio/chat): stop stream when trashing a thread from sidebar by @rolandtannous in fix(studio/chat): stop stream when trashing a thread from sidebar #5067
Studio: Local profile customization in settings and sync sidebar identity by @Imagineer99 in Studio: Local profile customization in settings and sync sidebar identity #5088
Studio: Show LoRA live logs and update GGUF quant options by @Imagineer99 in Studio: Show LoRA live logs and update GGUF quant options #5058
Studio: prefer mainstream clipboard copy over deprecated one by @G07cha in Studio: prefer mainstream clipboard copy over deprecated one #5109
Studio: Improve chat composition, fix scroll behaviour, and refine sidebar UX by @Imagineer99 in Studio: Improve chat composition, fix scroll behaviour, and refine sidebar UX #5089
Studio: forward standard OpenAI tools / tool_choice on /v1/responses (Codex compat) by @rolandtannous in Studio: forward standard OpenAI tools / tool_choice on /v1/responses (Codex compat) #5122
Studio: support images on /v1/messages (Anthropic-compat) by @rolandtannous in Studio: support images on /v1/messages (Anthropic-compat) #5128
Coerce TRL's tuple-cached _*_available flags to bool by @danielhanchen in Coerce TRL's tuple-cached _*_available flags to bool #5129
Studio: Smoother thread switching in chat by @Imagineer99 in Studio: Smoother thread switching in chat #5126
Studio: Replace assistant UI shared autoscroll with per-panel scrolling by @Imagineer99 in Studio: Replace assistant UI shared autoscroll with per-panel scrolling #5127
[pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in [pre-commit.ci] pre-commit autoupdate #5117
Fix tokenizer save gemma by @Datta0 in Fix tokenizer save gemma #5115
update gema4 chat templates by @Datta0 in update gema4 chat templates #5116
Bump installer floor to 2026.4.7 by @danielhanchen in Bump installer floor to 2026.4.7 #5134
fix/llamacpp_prebuilt_install by @mmathew23 in fix/llamacpp_prebuilt_install #5135
Studio: fix stale test_exception_result_cached test for vision cache by @danielhanchen in Studio: fix stale test_exception_result_cached test for vision cache #5145
fix: patch CONTROL type for special tokens in sentencepiece GGUF export by @octo-patch in fix: patch CONTROL type for special tokens in sentencepiece GGUF export #5080
fix(install): clear STUDIO_LOCAL_* env on POSIX normal install by @danielhanchen in fix(install): clear STUDIO_LOCAL_* env on POSIX normal install #5146
Add tauri by @wasimysaid in Add tauri #5144
Studio: detect reasoning_effort and preserve_thinking in chat templates by @danielhanchen in Studio: detect reasoning_effort and preserve_thinking in chat templates #5149

New Contributors

@cheehook made their first contribution in Fix Mistral DPO/preference training crash on non-xformers platforms (e.g. Intel XPU) #4889
@Ricardo-M-L made their first contribution in fix: check find() return value before adding offset in try_fix_tokenizer #4923
@Barath19 made their first contribution in fix(studio): default chart view to full training history #5007
@TF-MTGE made their first contribution in fix(studio): remove 300s cap on load_checkpoint (inherits 3600s default) #4922
@edamamez made their first contribution in Fix bitsandbytes ROCm install by using pip instead of uv #4966
@DavidSolanas made their first contribution in fix: prevent offline freeze by fixing stats retry and forwarding local_files_only #5016
@jonahsamost made their first contribution in grpo_compute_loss_slow called with wrong positional args #4887
@kimimgo made their first contribution in BUG: fix _fix_chat_template for ChatML templates missing add_generation_prompt (#4150) #4426
@CodeMan62 made their first contribution in fix audio dataset preview and finetuning #5043
@G07cha made their first contribution in Studio: prefer mainstream clipboard copy over deprecated one #5109
@octo-patch made their first contribution in fix: patch CONTROL type for special tokens in sentencepiece GGUF export #5080

Full Changelog: v0.1.36-beta...v0.1.37-beta

This discussion was created from the release Brand New UI + Qwen3.6.

Replies: 1 comment

Tom-Neverwinter
Apr 25, 2026

coding ide when

0 replies

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment