COMMITS
August 21, 2026
N
Add MLX disk cache control (#369)
Neil Mehta committed
N
Increase autofitted context length by dynamically adjusting prefill step size (#367)
Neil Mehta committed
August 19, 2026
N
Fix high memory usage during gemma4 image prefill (#362)
Neil Mehta committed
August 17, 2026
N
Add support for Muse Glimmer and its tool format (#364)
Neil Mehta committed
August 10, 2026
N
Add experimental server with chat completions endpoint (#353)
Neil Mehta committed
August 6, 2026
S
Merge pull request #361 from lmstudio-ai/samir/gemma4-fix
Samir committed
August 5, 2026
S
Match mlx-vlm Gemma 4 prefill logits
samir-lms committed
July 31, 2026
N
Add MLX context AutoFit load toggle (#355)
Neil Mehta committed
July 24, 2026
N
Enable disk-backed KV cache for text-only models (#352)
Neil Mehta committed
July 22, 2026
N
Mask invalid native tool bridge tokens (#351)
Neil Mehta committed
N
Update mlx dependencies (#350)
Neil Mehta committed
July 15, 2026
N
Update disk cache budget after auto-fit (#348)
Neil Mehta committed
July 14, 2026
N
Qwen 3.5 tool call grammar (#346)
Neil Mehta committed
N
Context length auto fit (#345)
Neil Mehta committed
July 13, 2026
N
Gemma4 tool calling grammar (#344)
Neil Mehta committed
June 24, 2026
N
Handle Gemma4 bidirectional visual prefill (#340)
Neil Mehta committed
June 15, 2026
N
Disable Qwen ragged attention kernel (#338)
Neil Mehta committed
June 11, 2026
N
Add Gemma 12b Unified Support (#334)
Neil Mehta committed
June 3, 2026
N
Clear Qwen text rope state before VLM prefill (#333)
Neil Mehta committed
May 28, 2026
N
Add disk-based caching and continuous batching for VLMs (#326)
Neil Mehta committed
April 20, 2026
M
Sync Qwen3.5 vision path with current mlx-vlm (#317)
Matt Clayton committed
April 14, 2026
N
Update gemma-4 test (#311)
Neil Mehta committed
April 13, 2026
N
Add vision feature caching for unified models (#309)
Neil Mehta committed
April 10, 2026
N
Add prompt caching checkpoints for sequential generation (#308)
Neil Mehta committed
April 8, 2026
N
Add unified arch for gemma4 (#305)
Neil Mehta committed
April 7, 2026
N
Upgrade mlx-lm and update to use new BatchedGeneration API (#304)
Neil Mehta committed
April 6, 2026
N
update mlx-vlm (#303)
Neil Mehta committed
April 3, 2026
N
Update requirements.txt; raise ValueError for unsupported model (#302)
Neil Mehta committed
March 31, 2026
W
Qwen 3.5 Unified (#298)
will-lms committed
March 25, 2026
W
Add prefill_step_size as load param (#295)
will-lms committed