Search results

qwen 3.5

Status: Open.

#994 In ml-explore/mlx-lm;

· Pall13 opened on Mar 13, 2026

Multimodal Inference Engine (similar to vLLM‑Omni) for MLX

Status: Open.

#993 In ml-explore/mlx-lm;

· Revive-Curiosity opened on Mar 13, 2026

Tool calling not detected for models with multi-token delimiters (e.g. Trinity)

Status: Open.

#984 In ml-explore/mlx-lm;

· juliensimon opened on Mar 11, 2026

BatchRotatingKVCache.merge() crashes on concurrent requests with different prompt lengths

Status: Open.

#983 In ml-explore/mlx-lm;

· juliensimon opened on Mar 11, 2026

Prefix cache reuse is broken for all hybrid-architecture models (sliding window, SSM/Mamba)

Status: Open.

#980 In ml-explore/mlx-lm;

· skcadri opened on Mar 11, 2026

Add ParoQuant (Pairwise Rotation Quantization) Support to MLX-LM

Status: Open.

#977 In ml-explore/mlx-lm;

· Revive-Curiosity opened on Mar 10, 2026

Strange cache behavior with 0.31.0 in server mode

Status: Open.

#975 In ml-explore/mlx-lm;

· putnam opened on Mar 9, 2026

📋 Documentation Enhancement Suggestion

Status: Open.

#972 In ml-explore/mlx-lm;

· croviatrust opened on Mar 9, 2026

load_model() should accept pre-loaded weights and config

Status: Open.

#969 In ml-explore/mlx-lm;

· blightbow opened on Mar 8, 2026

Discussion: Fuse gate/up expert projections via sanitize for MoE inference speedup

Status: Open.

#956 In ml-explore/mlx-lm;

· BurntToastGPT opened on Mar 7, 2026

Documentation for mlx_lm share — setup, performance, troubleshooting

Status: Open.

#955 In ml-explore/mlx-lm;

· guruswami-ai opened on Mar 6, 2026

[Feature]: FlashAttention on macOS

Status: Open.

#944 In ml-explore/mlx-lm;

· qdrddr opened on Mar 2, 2026