Liquid AI · Hugging Face
LFM2.5 1.2B
small machines
1,2B
Parameters
0,7 Go
Q4_K_M file
≥ 2,2 Go
Recommended VRAM
128 000
Max context
Speed by graphics card
Simulate in the “Configurator” →Standard llama-bench run (prompt 512 / generation 128 tokens), model fully in VRAM. Click a card for details. Electricity cost at the default price (0,25 €/kWh).
| Card · Q4_K_M | VRAM | Generation | Prompt | VRAM used | Power | € elec. / M tok | Price |
|---|---|---|---|---|---|---|---|
| RTX 4070 Ti | 12 Go | 473 tok/s | 31 828 | 1,4 Go | 81 W | 0,01 €/M tok | — |
Not measured on these cards yet — click for an estimate:
Technical sheet
- Published
- 2026-01-06
- Architecture
- 16 layers · 8 KV heads · dimension 64
- Attention
- Hybrid: 6 of 16 layers use standard attention, the others a fixed-memory mechanism (Mamba, linear attention) — a very light cache.
- Context cache (f16)
- ≈ 12 MB per 1,000 tokens · 0,4 Go for 32,768 tokens
- GGUF file
- bartowski/LiquidAI_LFM2.5-1.2B-Instruct-GGUF · LiquidAI_LFM2.5-1.2B-Instruct-Q4_K_M.gguf Download (0,7 Go)
- Official repo
- LiquidAI/LFM2.5-1.2B-Instruct
Public Hugging Face data (official config.json, GGUF repo). The context cache only counts full-attention layers.