Liquid AI · Hugging Face

LFM2.5 2.6B

small machines

2,7B

Parameters

1,6 Go

Q4_K_M file

≥ 3,1 Go

Recommended VRAM

131 072

Max context

Not measured yet

No card has run the TokenBench protocol on this model yet. The current protocol covers a selection of models; benchmarking any catalog model from the app is coming soon.

Measure my card →

Which cards should it fit on?

Cards on the site whose video memory covers the recommended 3,1 Go (file + 4,000-token context). An estimate, not a measurement. Click a card for the detailed estimate.

Technical sheet

Published: 2026-07-28
Architecture: 30 layers · 8 KV heads · dimension 64
Attention: Hybrid: 8 of 30 layers use standard attention, the others a fixed-memory mechanism (Mamba, linear attention) — a very light cache.
Context cache (f16): ≈ 16 MB per 1,000 tokens · 0,5 Go for 32,768 tokens
GGUF file: bartowski/LiquidAI_LFM2.5-2.6B-GGUF · LiquidAI_LFM2.5-2.6B-Q4_K_M.gguf Download (1,6 Go)
Official repo: LiquidAI/LFM2.5-2.6B

Public Hugging Face data (official config.json, GGUF repo). The context cache only counts full-attention layers.