whatmodel.sh / LFM2.5 2.6B

LFM2.5 2.6B

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

LFM2.5 2.6B is an open-weights language model from Liquid AI, released July 2026. Its published size is 2.7B parameters and a 131k-token context window. Licence: lfm1.0.

Catalogue identity notes

Liquid AI's LFM2.5-2.6B (LiquidAI/LFM2.5-2.6B @654f946, created 2026-07-28, LFM Open License v1.0), post-trained for agentic workloads. Scored through OpenBMB's internally reproduced 'LFM2.5-2.6B' column in the MiniCPM5-2B card (independent). Its own card's comparison table is held for the owner (no statement that Liquid ran the comparators).

Model information sources

Release date ↗ · Weights licence ↗

HF safetensors metadata 2,697,198,592 parameters (card names it 2.6B).

52.5±1.0#308 of 719 LLMsfrom 27 benchmark results across 2 sources

$ how it compares

The LLMs ranked either side of LFM2.5 2.6B, out of 828 with a score, then more from Liquid AI.

#modelscorewhereRAMreleased
368Qwen3-VL 4B Instruct Alibaba53±1local~4.1GB2025-10
369Claude 3 Opus Anthropic53±0comparison only2024-02
370LFM2.5 2.6B Liquid AI53±1local~3.2GB2026-07
371Bielik 11B v2.1 Instruct SpeakLeash53±3local~8.5GB2024-08
372Llama 3.1 70B Instruct Meta52±1api $0.72/M +local~27GB2024-07
more from Liquid AI
375LFM2.5 8B-A1B Liquid AI52±1local~6.5GB2026-05
398LFM2 24B-A2B Liquid AI51±1local~16GB2026-02
433LFM2.5 1.2B Thinking Liquid AI49±2local~2.2GB2026-01

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #163 of 300 · 16% of its score · 6 tests

Beats o1 (2024-12-17), o3-mini, DeepSeek R1 0528 here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Mini, Jamba 1.5 Large here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
BFCL V4 (December 2025 archive) +161.1% ✎▄·█·▄▄▄▄·▄·▄ beats 7/8medium · avg 43%▇▇ 5.6%at its level · several runs
BrowseComp-ZH9.8% ✎▄▄·▄▄▄·█···█ beats 3/8very hard · avg 13%▇ 2.6%typical
SWE-bench Verified6.0% ✎▄··▄▄█··▄·▄█ beats 4/8very hard · avg 12%▇ 2.5%sharp · noisy
τ²-bench Telecom90.4% ✎▁·▅█▆▅▂▃▃▂▃▁ beats 68/89hard · avg 32%▇ 1.9%sharp · at its level · noisy · small run
SWE-bench Pro (split unspecified)0.6% ✎█▂···▂···▄·▂ beats 3/8very hard · avg 4.3%▇ 1.7%sharp · too hard for it · noisy
τ³-bench Banking lab5.7% ✎▂▂·▅█▂▂▃▂▅▂▃ beats 7/20very hard · avg 4.0%▇ 1.6%blunt · too hard for it
coding▁▁▁▃▆██▇▆▄▃▁ #351 of 458 · 22% of its score · 5 tests

Trails Phi-3-mini-128k-instruct (Apr 2024), Code Tulu 2 13B, DeepSeek LLM 67B Chat here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LiveCodeBench v6 (window unspecified) +142.1% ✎▃····█·▃▅▃·▃ beats 2/8medium · avg 39%▇▇▇▇ 10%at its level · several runs
LiveCodeBench Pro 2025Q2 (Easy)30.9% ✎▄···█▄··▄█▄▄ beats 3/8hard · avg 33%▇▇ 5.0%at its level
LiveCodeBench Pro 2025Q2 (Medium)0.0% ✎█···▂··▄▂··▂ beats 3/8very hard · avg 2.0%▇ 3.0%sharp · too hard for it · noisy
SWE-bench Verified6.0% ✎▄··▄▄█··▄·▄█ beats 4/8very hard · avg 12%▇ 2.5%sharp · noisy
SWE-bench Pro (split unspecified)0.6% ✎█▂···▂···▄·▂ beats 3/8very hard · avg 4.3%▇ 1.7%sharp · too hard for it · noisy
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #114 of 393 * · 0.3% of its score · 1 test

Beats LongCat 2.0 (1.6T-A48B), Step 3.5 Flash (196B-A11B), GPT-5 mini here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
IFEval (aggregation unspecified)93.4% ✎▄▄····▄··███ beats 7/8easy · avg 77%▇ 0.3%noisy
math▁▂▃▅▆▅▄█▄▁▁▁ #233 of 398 · 39% of its score · 4 tests

Beats GPT-4o (2024-08-06), Command A (03-2025), GPT-4o (2024-11-20) here, though they rank higher overall. Trails OpenThinker 7B, Olmo 3 7B Instruct, Tulu 2 70B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
AIME 2025 +141.9% ✎█▄▄▄▄····█·▄ beats 3/8medium · avg 42%▇▇▇▇▇▇ 14%sharp · at its level · several runs
HMMT February 202633.7% ✎▅··▃▃·▃··▃·█ beats 3/8hard · avg 30%▇▇▇▇ 11%sharp · at its level
AIME 202645.2% ✎▄·▄█·▄▄···█▄ beats 3/8medium · avg 48%▇▇▇ 8.3%sharp · at its level
MATH-50089.6% ✎██▄▄▄··▄···▄ beats 3/8very easy · avg 89%▇▇ 5.2%typical
web search█·█▄██·▄▄·▄▄ #10 of 13 * · 2.6% of its score · 1 test

Beats Nemotron 3 Nano 4B, Gemma 4 E4B here, though they rank higher overall. Trails Qwen3.5 2B, LFM2.5 8B-A1B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
BrowseComp-ZH9.8% ✎▄▄·▄▄▄·█···█ beats 3/8very hard · avg 13%▇ 2.6%typical
other tests8 not in a task · 27% of its score
testresultvs the others in its rundifficultycountswhy
MMLU-Pro65.2% ✎▂··▂█·▂▂···▂ beats 3/8medium · avg 59%▇▇ 5.1%at its level
MMLU-Redux80.0% ✎▂···█··▂▂··▂ beats 5/8easy · avg 77%▇▇ 4.6%at its level
IFBench (aggregation unspecified) +159.0% ✎▅····▃▃·█·▃▃ beats 6/8medium · avg 39%▇▇ 4.4%at its level · several runs
OJBench11.2% ✎▃····▃▅··█▃▃ beats 2/8very hard · avg 12%▇ 3.2%typical
SuperGPQA26.2% ✎▃·▃·▃▅█····▃ beats 0/8hard · avg 32%▇ 2.9%at its level
LongBench v230.3% ✎▃······▅▅▃·█ beats 1/8very hard · avg 32%▇ 2.5%typical
Multi-IF (aggregation unspecified) +176.8% ✎▄▄···▄·▄·▄██ beats 8/8medium · avg 60%▇ 2.2%blunt · at its level · several runs
Claw-Eval average score lab62.8% ✎▄·▄··█▄▄▄▄▄▄ beats 3/9medium · avg 53%▇ 1.8%blunt · at its level

$ where to run it

llama.cpp · recommended

version

  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 2.6B · needs ~1.67 GB · from huggingface.co/LiquidAI (official)
    chat in the terminal
    llama cli -hf LiquidAI/LFM2.5-2.6B-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf LiquidAI/LFM2.5-2.6B-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 2.6B · needs ~1.67 GB · from huggingface.co/LiquidAI (official)

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/LiquidAI/LFM2.5-2.6B-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run LFM2.5 2.6B · needs ~1.67 GB · from huggingface.co/LiquidAI (official)
    chat in the terminal
    ollama run hf.co/LiquidAI/LFM2.5-2.6B-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 2.6B · needs ~1.6 GB, Mac only · from huggingface.co/LiquidAI (official)
    chat in the terminal
    mlx_lm.chat --model LiquidAI/LFM2.5-2.6B-MLX-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model LiquidAI/LFM2.5-2.6B-MLX-4bit

Official weights and documentation: huggingface.co/LiquidAI ↗

$ how the score was worked out · openbmb/MiniCPM5-2B model card 81%, LiquidAI/LFM2.5-2.6B model ca… 18%

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
openbmb/MiniCPM5-2B model card20▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 81%Apache-2.0
LiquidAI/LFM2.5-2.6B model card · lab-reported7▇▇▇▇▇ 18%LFM Open License v1.0
prior · pull toward the middle▇ 0%
weighted average= 52.5

$ 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