whatmodel.sh / LFM2.5 1.2B Instruct

LFM2.5 1.2B Instruct

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

LFM2.5 1.2B Instruct is an open-weights language model from Liquid AI, released January 2026. Its published size is 1.17B parameters and a 33k-token context window. Licence: lfm1.0.

Catalogue identity notes

LiquidAI/LFM2.5-1.2B-Instruct @0f604ad (created 2026-01-06, licence lfm1.0). Liquid AI LFM2.5 1.2B instruction model (Jan 2026), sibling of the catalogued LFM2.5-1.2B-Thinking. Lab rows from the LFM2.5-1.2B-Thinking card table (existing runs with Qwen3-1.7B, Granite 4.0 1B, Gemma 3 1B, Llama 3.2 1B); independent EuroEval rows.

Model information sources

Release date ↗ · Weights licence ↗

HF safetensors metadata 1,170,340,608 parameters (LiquidAI/LFM2.5-1.2B-Instruct @0f604ad).

43.5±1.8#475 of 719 LLMsfrom 10 benchmark results across 2 sources

$ how it compares

The LLMs ranked either side of LFM2.5 1.2B Instruct, out of 828 with a score, then more from Liquid AI.

#modelscorewhereRAMreleased
562Code Llama 70B Python Meta44±1local~28GB2024-01
563Llama 3.1 8B (base) Meta44±3local~6.2GB2024-07
564LFM2.5 1.2B Instruct Liquid AI44±2local~2.2GB2026-01
565Granite 3.0 8B Instruct IBM43±1local~6.6GB2024-10
566Phi-3.5-mini-instruct Microsoft43±1local~3.7GB2024-08
more from Liquid AI
370LFM2.5 2.6B Liquid AI53±1local~3.2GB2026-07
375LFM2.5 8B-A1B Liquid AI52±1local~6.5GB2026-05
398LFM2 24B-A2B Liquid AI51±1local~16GB2026-02

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #247 of 300 * · 3.1% of its score · 1 test

Beats DeepSeek R1 0528 Qwen3 8B, Magistral Small 1.0 (2506), Llama 4 Scout (109B-A17B) here, though they rank higher overall. Trails Granite 4.0 H 1B, Granite 4.0 1B, SmolLM3 3B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
BFCL v3 lab49.1% ✎▄▄·······▄██ beats 2/6medium · avg 56%▇ 3.1%blunt · at its level
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #161 of 393 * · 10% of its score · 1 test

Beats MiniMax-M2, MiniCPM5 2B, Granite 4.2 30B here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
IFEval mean of four measures lab86.2% ✎█·█··█·██·██ beats 5/6very easy · avg 87%▇▇▇▇ 10%typical
math▁▂▃▅▆▅▄█▄▁▁▁ #299 of 398 · 43% of its score · 3 tests

Beats GPT-4 (0613), Athene-70B, Gemma 2 27B IT here, though they rank higher overall. Trails Gemma 7B (base), DeepSeek R1 Distill Qwen 1.5B, CodeGemma 7B (v1.0) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
GSM8K lab64.5% ✎█····▄▄▄···█ beats 2/6very easy · avg 87%▇▇▇▇▇▇▇▇▇ 24%sharp
AIME 2025 lab14.0% ✎▄█··▄···▄··█ beats 4/6hard · avg 33%▇▇▇▇ 11%sharp · noisy
MATH-500 lab63.2% ✎▄··█▄·▄···▄▄ beats 4/6easy · avg 80%▇▇▇ 8.7%at its level
other tests5 not in a task · 42% of its score
testresultvs the others in its rundifficultycountswhy
MMLU-Pro lab44.4% ✎█···█·██·███ beats 4/6medium · avg 54%▇▇▇▇▇ 12%at its level
GPQA Diamond lab38.9% ✎█··········█ beats 5/5hard · avg 41%▇▇▇▇▇ 12%sharp · noisy
IFBench (aggregation unspecified) lab47.3% ✎▄·█·█······█ beats 6/6medium · avg 43%▇▇▇▇ 10%at its level
EuroEval HellaSwag (English), test split, 5-shot generative MC, accuracy32.5% ✎▃·▁▁▃▁▂▅█▇▄▁ beats 15/100medium · avg 68%▇▇ 4.8%at its level · noisy
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy61.2% ✎▁▁▁▁▁▁▁▃▄█▄▃ beats 31/146easy · avg 83%▇▇ 4.0%at its level · noisy

$ where to run it

llama.cpp · recommended

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run LFM2.5 1.2B Instruct · needs ~0.73 GB · from huggingface.co/LiquidAI (official)
    chat in the terminal
    llama cli -hf LiquidAI/LFM2.5-1.2B-Instruct-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf LiquidAI/LFM2.5-1.2B-Instruct-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run LFM2.5 1.2B Instruct · needs ~0.73 GB · from huggingface.co/LiquidAI (official)

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/LiquidAI/LFM2.5-1.2B-Instruct-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run LFM2.5 1.2B Instruct · needs ~0.73 GB · from huggingface.co/LiquidAI (official)
    chat in the terminal
    ollama run hf.co/LiquidAI/LFM2.5-1.2B-Instruct-GGUF:Q4_K_M

MLX (Mac)

version

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run LFM2.5 1.2B Instruct · needs ~0.7 GB, Mac only · from huggingface.co/LiquidAI (official)
    chat in the terminal
    mlx_lm.chat --model LiquidAI/LFM2.5-1.2B-Instruct-MLX-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model LiquidAI/LFM2.5-1.2B-Instruct-MLX-4bit

Official weights and documentation: huggingface.co/LiquidAI ↗

$ how the score was worked out · LiquidAI/LFM2.5-1.2B-Thinking… 90%, EuroEval leaderboards (result… 9%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
LiquidAI/LFM2.5-1.2B-Thinking model card · lab-reported8▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 90%LFM Open License v1.0
EuroEval leaderboards (results collection), English generative leaderboard2▇▇▇ 9%MIT
prior · pull toward the middle▇ 1%
weighted average= 43.5

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json