LFM2.5-8B-A1B-GGUF
LFM2 is a new generation of hybrid models developed by Liquid AI, specifically designed for edge AI and on-device deployment. It sets a new standard in terms of quality, speed, and memory efficiency.
Params
—
Context
—
Downloads 30d
563 K
Likes
300
Download history
daily snapshots · 31 days
▲ 377 K in the last 30 days (201.5%)
563 K197 K
Aug 22Sep 1Sep 11Sep 20
563 K187 K
Aug 21Aug 31Sep 10Sep 20
Can you run it?
Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.
| File | Quant | Size | Est. VRAM | Verdict on RTX 4090 · 24 GB |
|---|---|---|---|---|
| LFM2.5-8B-A1B-Q4_0.gguf | Q4_0 | 4.8 GB | 5.8 GB | ✅ Runs comfortably |
| LFM2.5-8B-A1B-Q4_K_M.gguf | Q4_K_M | 5.2 GB | 6.2 GB | ✅ Runs comfortably |
| LFM2.5-8B-A1B-Q5_K_M.gguf | Q5_K_M | 6.0 GB | 7.1 GB | ✅ Runs comfortably |
| LFM2.5-8B-A1B-Q6_K.gguf | Q6_K | 7.0 GB | 8.2 GB | ✅ Runs comfortably |
| LFM2.5-8B-A1B-Q8_0.gguf | Q8_0 | 9.0 GB | 10.4 GB | ✅ Runs comfortably |
| LFM2.5-8B-A1B-BF16.gguf | GGUF | 16.9 GB | 19.1 GB | ✅ Runs comfortably |
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.
Run it
copy-paste, exact tags checked against the Hub$ ollama run lfm2-5-8b-a1b-gguf # pin the quantization explicitly $ ollama run lfm2-5-8b-a1b-gguf-q4_0
$ huggingface-cli download LiquidAI/LFM2.5-8B-A1B-GGUF-GGUF \
LFM2.5-8B-A1B-Q4_0.gguf --local-dir .
$ llama-cli -m LFM2.5-8B-A1B-Q4_0.gguf \
-c 8192 -ngl 99 -t 8 --color
$ curl -s https://aimodelscomparison.com/api/v1/models/lfm2-5-8b-a1b-gguf
{
"hf_id": "LiquidAI/LFM2.5-8B-A1B-GGUF",
"params_b": null,
"context_length": null,
"license": { "id": "other", "commercial": "unknown" },
"downloads_30d": 563450,
"vram_estimates": [
{ "quant": "GGUF", "gb": 19.1 },
{ "quant": "Q4_0", "gb": 5.8 }
],
"updated_at": "2026-08-25T01:00:57Z"
}
Specifications
- Licence
- other
- First seen on the Hub
- 2026-05-24
- Base model
- LFM2.5-8B-A1B
- Training datasets
- undisclosed
- Added to our catalog
- 2026-08-21
Family
Base model and the most-downloaded derivatives in the catalog.
Compare with any text-generation model
Appears in