[whatmodel.sh](../../README.md) / LFM2 2.6B
# LFM2 2.6B
Liquid AI · released 2025-09 · 2.57B params · lfm1.0
**score 44.2** ±1.4 · #493 of 732 · from 13 benchmark results across 2 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks LFM2 2.6B among every model on that task or test, worst to best.
task vs the others rank counts tests
chat & writing ▁▁▁▃▆▄●█▆▇▇▃ #250 of 402* 12% 1
math ▁▁▃▅●▆▄█▄▂▁▁ #298 of 431* 8.2% 2
chat & writing · #250 of 402 · 12% of its score
Beats [Mistral Small 4 (119B-A6.5B)](mistral-small-4.md), [Qwen2.5 Coder 32B Instruct](qwen2-5-coder-32b.md), [Ministral 3 14B Reasoning 2512](ministral-3-14b-reasoning.md) here, though they rank higher overall. Trails [Flan-T5 XXL](flan-t5-xxl.md), [Granite 4.0 H 1B](granite-4-0-h-1b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
IFEval mean of four measures (mak… 79.6% █··▄▄·●····▄ beats 4/5 easy · avg 84% 12% typical
math · #298 of 431 · 8.2% of its score
Beats [Molmo 2 8B](molmo-2-8b.md), [DeepSeek-Coder-V2 Instruct (0724)](deepseek-coder-v2-0724.md), [Llama 3.1 Nemotron 70B Instruct](llama-3-1-nemotron-70b.md) here, though they rank higher overall. Trails [DeepSeek R1 Distill Qwen 1.5B](deepseek-r1-distill-qwen-1-5b.md), [Gemma 7B (base)](gemma-7b-base.md), [Mistral 7B v0.1 (base)](mistral-7b-v0-1.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
MATH-500 (maker) 63.6% ▃····●·█···▃ beats 1/5 easy · avg 79% 7.1% at its level
GSM8K loose substring accuracy (m… 82.4% █·█··█●█···█ beats 3/5 easy · avg 83% 1.1% blunt
other tests (10)
test result vs the others in its run difficulty counts why
MMLU (maker) 64.4% ▄█···●·····▄ beats 3/5 medium · avg 73% 18% at its level
GPQA Diamond (maker) 26.6% █●····███··█ beats 1/5 very hard · avg 34% 11% sharp · too hard for it
MMLU-Pro (maker) 26.0% ██●···██···█ beats 2/5 medium · avg 42% 10% typical
IFBench (aggregation unspecified)… 22.2% █··█●█··█··█ beats 2/5 hard · avg 31% 9.4% typical
Multi-IF three-turn average (make… 60.3% ▄··█●··▄···▄ beats 3/5 easy · avg 70% 8.6% at its level
MMMLU seven languages (Liquid) (m… 55.4% ▄·█····●···▄ beats 4/5 medium · avg 59% 7.8% at its level
MGSM six languages, loose (Liquid… 74.3% ▄·▄·●···▄··▄ beats 3/5 easy · avg 81% 5.8% at its level
EuroEval HellaSwag (English), tes… 64.5% ▃·▁▁▂▁▂●█▇▄▁ beats 39/100 medium · avg 67% 3.8% sharp · at its level · noisy
EuroEval Life in the UK (English … 76.3% ▁▁▁▁▁▁▁▂●█▄▃ beats 54/146 easy · avg 82% 3.1% at its level · noisy
GSM-Plus (maker) 60.8% █·····██●·██ beats 3/5 easy · avg 71% 2.0% blunt · at its level
source results share of the score
LFM2 Technical Report (submitted / make… 11 92% ▇▇▇▇▇▇▇▇▇▇▇
EuroEval leaderboards 2 7% ▇
prior: pull toward the middle 1% ▇
weighted average = 44.2
Licences: LFM2 Technical Report ([CC BY 4.0](../../licenses/cards-gap-NOTICE.txt)) · EuroEval leaderboards ([MIT](https://github.com/EuroEval/leaderboards/blob/1a77f1d2f8ba1ba05ab73287fb090d76bf41b2e0/LICENSE))
## How to run it
### llama.cpp (recommended)
1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md)
```sh
# macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
# Windows
irm https://llama.app/install.ps1 | iex
```
2. Run LFM2 2.6B (Q4_K_M, needs ~1.56 GB, from [huggingface.co/LiquidAI](https://huggingface.co/LiquidAI/LFM2-2.6B-GGUF/tree/a759abdc5955d4ca97763e5cb7ff3940589ba898))
```sh
llama cli -hf LiquidAI/LFM2-2.6B-GGUF:Q4_K_M
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf LiquidAI/LFM2-2.6B-GGUF:Q4_K_M
```
### LM Studio
1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download)
2. Run LFM2 2.6B (Q4_K_M, needs ~1.56 GB, from [huggingface.co/LiquidAI](https://huggingface.co/LiquidAI/LFM2-2.6B-GGUF/tree/a759abdc5955d4ca97763e5cb7ff3940589ba898))
```sh
lms get https://huggingface.co/LiquidAI/LFM2-2.6B-GGUF@Q4_K_M -y
```
3. Then run lms ls to see its name, and lms chat followed by that name
Other versions: 4bit (~1.4 GB, Mac only) `lms get https://huggingface.co/mlx-community/LFM2-2.6B-4bit -y` · 8bit (~2.7 GB, Mac only) `lms get https://huggingface.co/mlx-community/LFM2-2.6B-8bit -y`
### Ollama
1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download)
```sh
# Linux
curl -fsSL https://ollama.com/install.sh | sh
```
2. Run LFM2 2.6B (Q4_K_M, needs ~1.56 GB, from [huggingface.co/LiquidAI](https://huggingface.co/LiquidAI/LFM2-2.6B-GGUF/tree/a759abdc5955d4ca97763e5cb7ff3940589ba898))
```sh
ollama run hf.co/LiquidAI/LFM2-2.6B-GGUF:Q4_K_M
```
### MLX (Mac)
1. Install MLX (Mac). Apple silicon Macs only, text only.
```sh
# with uv
uv tool install mlx-lm
# or
pip install mlx-lm
```
2. Run LFM2 2.6B (4bit, needs ~1.4 GB, Mac only, from [huggingface.co/mlx-community](https://huggingface.co/mlx-community/LFM2-2.6B-4bit/tree/493071ebca3592c63085e19612ded87192a2a0dd))
```sh
mlx_lm.chat --model mlx-community/LFM2-2.6B-4bit
```
3. Or start an OpenAI-compatible server at http://localhost:8080/v1
```sh
mlx_lm.server --model mlx-community/LFM2-2.6B-4bit
```
Other versions: 8bit (~2.7 GB, Mac only) `mlx_lm.chat --model mlx-community/LFM2-2.6B-8bit`
Model repository and documentation: [huggingface.co/LiquidAI](https://huggingface.co/LiquidAI/LFM2-2.6B)
[interactive version](https://whatmodel.sh/model/lfm2-2-6b) · [JSON](https://whatmodel.sh/api?format=json&all=1)