Search results

Qwen3-Coder-Next-4bit produces corrupted / garbage tokens with long prompts

Status: Open.

#856 In ml-explore/mlx-lm;

· cubist38 opened on Feb 7, 2026

mlx_lm.server crashes on Metal OOM instead of returning an HTTP error

Status: Open.

#854 In ml-explore/mlx-lm;

· BrunoCerberus opened on Feb 7, 2026

Support for structured output in batch mode?

bugSomething isn't workingSomething isn't working

Status: Open.

#852 In ml-explore/mlx-lm;

· do-me opened on Feb 6, 2026

n-gram hashing for speculative decoding

Status: Open.

#851 In ml-explore/mlx-lm;

· nightscape opened on Feb 6, 2026

Speculative decoding produces incorrect output with Qwen3 models (skips tokens)

Status: Open.

#846 In ml-explore/mlx-lm;

· janhilgard opened on Feb 5, 2026

Memory during training (LoRA)

Status: Open.

#828 In ml-explore/mlx-lm;

· simap opened on Jan 30, 2026

[Improvement] Enhance <code>mlx_lm.chat</code> CLI experience (history, line editing, and formatting)

Status: Open.

#818 In ml-explore/mlx-lm;

· Vlor999 opened on Jan 28, 2026

Feature: Support OpenResponses API

Status: Open.

#815 In ml-explore/mlx-lm;

· njbrake opened on Jan 28, 2026

Sherry/Tequila Ternary "BitNet" Support

Status: Open.

#812 In ml-explore/mlx-lm;

· TomLucidor opened on Jan 27, 2026

[Feature Request] xLSTM

Status: Open.

#805 In ml-explore/mlx-lm;

· Anri-Lombard opened on Jan 24, 2026

MLX Ecosystem Maintenance Analysis - Easy Wins &amp; High-Impact Issues

Status: Open.

#803 In ml-explore/mlx-lm;

· jrp2014 opened on Jan 23, 2026

Problem with CompletionsDataset and mask-prompt

Status: Open.

#786 In ml-explore/mlx-lm;

· tomekkasp opened on Jan 20, 2026