LFM2 350M Parity 4bit MLX

Same-origin MLX artifact for mesh-llm backend comparison.

Source checkpoint: LiquidAI/LFM2-350M
Conversion path: original checkpoint -> MLX 4bit
Intended pair: meshllm/lfm2-350m-parity-q4_k_m-gguf

Validation

Validated locally with the mesh-llm exact smoke suite on 2026-04-06.

This pair is useful as a backend-drift detector rather than a clean parity canary. In the same-origin exact run, the MLX side materially outperformed the GGUF side on several simple prompts.

Files

MLX model directory contents

Downloads last month: 28

Safetensors

Model size

55.4M params

Tensor type

BF16

U32

MLX

Hardware compatibility

4-bit

Model tree for meshllm/lfm2-350m-parity-4bit-mlx

Base model

LiquidAI/LFM2-350M

Quantized

(32)

this model