Source: LiquidAI/LFM2.5-8B-A1B @ revision b9aebfcbe28b6cb374042f495d733037550ab146 (LFM Open License v1.0 — see LICENSE in this repo).
Quantized with mlx_lm.convert (mlx-lm 0.31.3): affine, 4-bit, group size 32.

Smoke gate

Before upload this pack passed a deterministic coherence gate: greedy 32-token chat generation loaded through mlx_lm.load, judged for emptiness, repetition loops, multi-script gibberish, and special-token debris. Verdict: ok.

Usage

bash

pip install mlx-lm
mlx_lm.generate --model majentik/LFM2.5-8B-A1B-MLX-4bit --prompt "Hello"

Evaluation

Benchmark	Score
arc_easy_acc	0.5050
hellaswag_acc	0.4500

License

LFM Open License v1.0 (lfm1.0): commercial use permitted with attribution. The full text ships as LICENSE in this repository, copied verbatim from the upstream model.

Available tiers

Repo	Quant
majentik/LFM2.5-8B-A1B-MLX-2bit	2bit
majentik/LFM2.5-8B-A1B-MLX-3bit	3bit
majentik/LFM2.5-8B-A1B-MLX-4bit	4bit
majentik/LFM2.5-8B-A1B-MLX-5bit	5bit
majentik/LFM2.5-8B-A1B-MLX-6bit	6bit
majentik/LFM2.5-8B-A1B-MLX-8bit	8bit
majentik/LFM2.5-8B-A1B-MLX-MXFP4	MXFP4