Skip to content

Pull requests: Blaizzy/mlx-vlm

New pull request New
60 Open 488 Closed
60 Open 488 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Attempt to fix Qwen VL generation
#1032 opened Apr 17, 2026 by pcuenca Contributor Loading…
2
Add KV cache quantization for continuous batching
#1030 opened Apr 17, 2026 by Blaizzy Owner Loading…
4 of 5 tasks
2
Add DFlash speculative decoding (single + batch + server)
#1029 opened Apr 16, 2026 by Blaizzy Owner Loading…
7 tasks done
Add vision feature caching to all models
#1028 opened Apr 16, 2026 by Blaizzy Owner Loading…
6 tasks done
Security: Remote image fetch lacks timeout and payload size limits (DoS risk)
#1026 opened Apr 16, 2026 by tomaioo Loading…
1
Expose presence_penalty, frequency_penalty, and per-penalty context_size on the server API
#1023 opened Apr 14, 2026 by esaruoho Loading…
1
refactor: improve model loading and resource handling in utils.py
#1019 opened Apr 13, 2026 by SyedaAnshrahGillani Loading…
Add Youtu-VL
#1018 opened Apr 13, 2026 by MollySophia Loading…
Feature: Add native Gemma 4 video support
#1017 opened Apr 12, 2026 by hybridherbst Loading…
fix: propagate the verbose to the Prefill tqdm
#1015 opened Apr 12, 2026 by PeterStaar-IBM Loading…
1
server: indicate finish reason properly when model made a tool call.
#1014 opened Apr 12, 2026 by viktike Contributor Loading…
Resolve no images crash for qwen3_vl and qwen3_vl_moe generate call
#1013 opened Apr 11, 2026 by urimem Loading…
perf: close 5.5% decode gap vs mlx_lm.server on streaming chat endpoint
#1012 opened Apr 11, 2026 by chilang Loading…
fix: use OpenAI chat-completion field names in /chat/completions usage
#1009 opened Apr 10, 2026 by chilang Loading…
fix: replace NaN from all-masked SDPA padding rows in Gemma 4 vision
#1006 opened Apr 10, 2026 by fabiopili Loading…
4 tasks done
feat: OpenAI Responses API with structured tool calling and multi-turn support
#996 opened Apr 9, 2026 by eloe Loading…
feat: prompt prefix caching with TTL eviction and TurboQuant support
#995 opened Apr 9, 2026 by eloe Loading…
1
feat: add logprobs support to /chat/completions
#994 opened Apr 9, 2026 by eloe Loading…
feat: add JSON mode via response_format parameter
#993 opened Apr 9, 2026 by eloe Loading…
feat: enforce tool_choice parameter in chat/completions
#992 opened Apr 9, 2026 by eloe Loading…
feat: add stop sequences support for both endpoints
#991 opened Apr 9, 2026 by eloe Loading…
fix: return finish_reason=tool_calls when tool calls detected
#990 opened Apr 9, 2026 by eloe Loading…
feat: concurrency guard for Metal GPU serialization
#989 opened Apr 9, 2026 by eloe Loading…
Fix sft_trainer crash when adapter_file is None (#908)
#987 opened Apr 9, 2026 by H-A-Khan Loading…
1
Fix LoRA scaling: divide alpha by rank (#845)
#986 opened Apr 9, 2026 by H-A-Khan Loading…
1
ProTip! Exclude everything labeled bug with -label:bug.