Search results

MTPLX gets stuck when used with Hermes Agent

Status: Open.

#86 In youssofal/MTPLX;

· benmarte opened on May 30, 2026

Feature Request: Native MTPLX conversion for Qwen3.6-35B-A3B (Q2_K_XL / IQ4_XS)

Status: Open.

#85 In youssofal/MTPLX;

· TheDevsGarden opened on May 25, 2026

ValueError: SparseDistribution probabilities must have positive mass during speculative drafting

Status: Open.

#82 In youssofal/MTPLX;

· Jonathangadeaharder opened on May 24, 2026

Feature request: supply --api-key via env var or file (avoid argv exposure with --host 0.0.0.0)

Status: Open.

#81 In youssofal/MTPLX;

· daitamu opened on May 24, 2026

_MaxIdleWatchdog sends requests to /health without auth headers, causing 401s and disabling fan control

Status: Open.

#80 In youssofal/MTPLX;

· shanemmattner opened on May 24, 2026

Implement RotorQuant/IsoQuant/PlaneQuant support

Status: Open.

#74 In youssofal/MTPLX;

· sangemaru opened on May 17, 2026

Support for PARO quantized models

Status: Open.

#73 In youssofal/MTPLX;

· sangemaru opened on May 17, 2026

Have any kv cache compress ?

Status: Open.

#72 In youssofal/MTPLX;

· tienlbhoc opened on May 17, 2026

In M1MAX 64g， the speed is same as mlx normal model,27b, 19t/s， use MTP 19t/s,

Status: Open.

#68 In youssofal/MTPLX;

· zwqjoy opened on May 16, 2026

Reproducing the 2.24× speedup: documented MLX patch alone doesn't get there — please publish the actual fork commit

Status: Open.

#64 In youssofal/MTPLX;

· benjamin-levin opened on May 14, 2026

OpenCode (Desktop and CLI) *extremely* slow

Status: Open.

#63 In youssofal/MTPLX;

· derekrodgers opened on May 13, 2026

Add PFlash/DFlash Support

Status: Open.

#56 In youssofal/MTPLX;

· TomLucidor opened on May 11, 2026