Search results

Unify LoRA adapter loading for text-only MLX-LM models

enhancementNew feature or requestNew feature or request

Status: Open.

#1184 In Blaizzy/mlx-vlm;

· Blaizzy opened on May 15, 2026

Qwen2.5-VL default max_pixels in preprocessor_config.json is 12× the training resolution; degrades grounding accuracy

bugSomething isn't workingSomething isn't working

Status: Open.

#1175 In Blaizzy/mlx-vlm;

· NivDvir opened on May 12, 2026

Support APC with KV-cache quantization (critical for TTFT & long Agent tasks for local-first AI)

Status: Open.

#1174 In Blaizzy/mlx-vlm;

· Edsuns opened on May 12, 2026

Support response_format type=json_object

Status: Open.

#1171 In Blaizzy/mlx-vlm;

· eloe opened on May 12, 2026

20260508 -- Diagnostics Report — 7 failure(s), 2 harness issue(s) (mlx-vlm 0.5.0)

Status: Open.

#1156 In Blaizzy/mlx-vlm;

· jrp2014 opened on May 8, 2026

Add tools / function-calling support to /v1/responses

Status: Open.

#1136 In Blaizzy/mlx-vlm;

· jasonvassallo opened on May 7, 2026

No model is included in the JSON response when making a request to http://127.0.0.1:8080/v1/models

Status: Open.

#1133 In Blaizzy/mlx-vlm;

· slushhh opened on May 6, 2026

mlx_vlm.server crashes if the ~/.cache/huggingface/hub directory is missing

Status: Open.

#1132 In Blaizzy/mlx-vlm;

· slushhh opened on May 6, 2026

Server generation exceptions can spin in a hot error loop

enhancementNew feature or requestNew feature or request

Status: Open.

#1107 In Blaizzy/mlx-vlm;

· eloe opened on May 5, 2026

2026-05-01 -- Diagnostics Report — 2 failure(s), 9 harness issue(s) (mlx-vlm 0.4.5+)

Status: Open.

#1100 In Blaizzy/mlx-vlm;

· jrp2014 opened on May 1, 2026

DFlash + concurrent requests: TypeError in rollback_speculative_cache (qwen3_5/language.py:523)

Status: Open.

#1086 In Blaizzy/mlx-vlm;

· nnorris7 opened on Apr 28, 2026

module 'mlx.core' has no attribute 'new_thread_local_stream'

Status: Open.

#1085 In Blaizzy/mlx-vlm;

· zhangcz-ai opened on Apr 28, 2026