LFM2 models
15 LFM2 models from Liquid AI on Hugging Face, from 230M to 8.5B parameters, published by Liquid AI in BF16, with contexts from 125K to 128K tokens. The smallest official model, LFM2.5-230M, needs about 0.5 GB of VRAM at its published precision; the cheapest live fit is RTX 3070 at $0.088/hr.
Pick a size
One row per official LFM2 size: the VRAM it needs at each precision, the cheapest GPU that holds it today, and the KV cache for a 32K-token context.
| Model | Parameters | Native VRAM | FP8 VRAM | INT4 VRAM | Live GPU fit (native) | Est. $/hr | KV cache at 32K |
|---|---|---|---|---|---|---|---|
| LFM2.5-230M | 230M | 0.5 GB | 0.3 GB | 0.1 GB | RTX 3070 | $0.088/hr | 0.38 GB |
| LFM2.5-350M | 354M | 0.8 GB | 0.4 GB | 0.2 GB | RTX 3070 | $0.088/hr | 0.38 GB |
| LFM2.5-1.2B-Instruct | 1.2B | 2.6 GB | 1.3 GB | 0.7 GB | RTX 3070 | $0.088/hr | 0.38 GB |
| LFM2.5-VL-1.6B | 1.6B | 3.6 GB | 1.8 GB | 0.9 GB | RTX 3070 | $0.088/hr | 0.38 GB |
| LFM2.5-2.6B | 2.7B | 6.0 GB | 3.0 GB | 1.5 GB | RTX 3070 | $0.088/hr | 0.50 GB |
| LFM2-8B-A1B | 8.3B | 18.6 GB | 9.3 GB | 4.7 GB | RTX A5000 | $0.176/hr | 0.38 GB |
| LFM2.5-8B-A1B | 8.5B | 18.9 GB | 9.5 GB | 4.7 GB | RTX A5000 | $0.176/hr | 0.38 GB |
VRAM is weights times a flat 1.2 overhead; the KV cache is a separate per-model figure at 16-bit, shown where the architecture is published. See the methodology.
Official models (13)
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| LFM2.5-230M | 230M | BF16 | 0.5 GB | RTX 3070 | $0.088/hr |
| LFM2.5-230M-Base | 230M | BF16 | 0.5 GB | RTX 3070 | $0.088/hr |
| LFM2.5-350M | 354M | BF16 | 0.8 GB | RTX 3070 | $0.088/hr |
| LFM2.5-350M-Base | 354M | BF16 | 0.8 GB | RTX 3070 | $0.088/hr |
| LFM2-1.2B | 1.2B | BF16 | 2.6 GB | RTX 3070 | $0.088/hr |
| LFM2.5-1.2B-Base | 1.2B | BF16 | 2.6 GB | RTX 3070 | $0.088/hr |
| LFM2.5-1.2B-Instruct | 1.2B | BF16 | 2.6 GB | RTX 3070 | $0.088/hr |
| LFM2.5-1.2B-JP-202606 | 1.2B | BF16 | 2.6 GB | RTX 3070 | $0.088/hr |
| LFM2.5-VL-1.6B | 1.6B | BF16 | 3.6 GB | RTX 3070 | $0.088/hr |
| LFM2.5-2.6B | 2.7B | BF16 | 6.0 GB | RTX 3070 | $0.088/hr |
| LFM2.5-2.6B-Base | 2.7B | BF16 | 6.0 GB | RTX 3070 | $0.088/hr |
| LFM2-8B-A1B | 8.3B | BF16 | 18.6 GB | RTX A5000 | $0.176/hr |
| LFM2.5-8B-A1B | 8.5B | BF16 | 18.9 GB | RTX A5000 | $0.176/hr |
Quantized versions (1)
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| LFM2-1.2B-MLX-bf16 | 1.2B | BF16 | 2.6 GB | RTX 3070 | $0.088/hr |
Fine-tunes and community models (1)
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| SauerkrautLM-GLiNEREL-314m | 314M | BF16 | 0.7 GB | RTX 3070 | $0.088/hr |
VRAM is for the precision the model is published in: the weights times a flat 1.2 overhead, with the KV cache not included. See the methodology. The fit shown is the lowest-priced single GPU type that holds the model at that precision, or the lowest-priced multi-GPU set (up to 8) when none does.
More from Liquid AI
- LiquidAI (5 models)