whatmodel.sh / LFM3 1.7B

LFM3 1.7B

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

LFM3 1.7B is an open-weights language model from Liquid AI, released October 2026. Its published size is 1.73B parameters and a 262k-token context window. Licence: lfm1.0.

Catalogue identity notes

Post-trained text checkpoint, distinct from Base. Requires llama.cpp b11960/v0.8.0 or newer. Memory estimate includes weight overhead, not maximum-context KV cache.

Model information sources

Release date ↗

35.0±15.0#2 of 2 LLMsfrom 7 benchmark results across 1 source

Score within the GPQA Diamond comparison set only.

$ how it compares

Scored LLMs released closest to LFM3 1.7B, out of 1192 with a score, then more from Liquid AI.

#modelscorewhereRAMreleased
3GPT-6.1 Terra OpenAI84±1api $2.00/M2026-10
4Gemini 4 Carbon Google83±1limited preview2026-10
7Muse Wonder Full Meta82±2Muse Code only2026-10
14Kimi K3.1 Pro Moonshot AI80±2api price unknown2026-10
18Muse Wonder Meta78±1api $8.00/M2026-10
more from Liquid AI
528LFM2.5 2.6B Liquid AI53±1local~2.3GB2026-07
539LFM2.5 8B-A1B Liquid AI52±1local~5.5GB2026-05
570LFM2 24B-A2B Liquid AI51±1local~14GB2026-02

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks█··········█ #2 of 2 * · 31% of its score · 6 tests
testresultvs the others in its rundifficultycountswhy
WebArena-Verified lab38.4% ✎█··█···█·█·█ beats 1/4hard · avg 32%▇▇▇▇▇▇▇▇▇▇▇ 27%sharp · small run
SWE-bench Verified lab43.4% ✎▁▁▁▁▁▁▂▅█▆▁▁ beats 33/136medium · avg 47%▇▇ 4.1%at its level · small run
SWE-bench Multilingual lab23.0% ✎▁··▃▃▁▄▇█▂▂▁ beats 1/63—not counted yet: no other model in the same run to compare it with
Terminal-Bench 2.1 lab12.9% ✎▁·▁·▁▃▄▃▆█▅█ beats 3/63—not counted yet: no other model in the same run to compare it with
Humanity's Last Exam with tools lab13.9% ✎▃▃▁▂▃▃▃█▇▇▆▂ beats 0/58—not counted yet: no other model in the same run to compare it with
MCP-Atlas (public set) lab32.6% ✎▂·▁▂▃·▁▃█▆▆▂ beats 5/39—not counted yet: no other model in the same run to compare it with
coding█··········█ #2 of 2 * · 10% of its score · 4 tests
testresultvs the others in its rundifficultycountswhy
SciCode (level unspecified) lab28.5% ✎▁····▃▄▄▅█▇▃ beats 10/38hard · avg 30%▇▇ 5.8%at its level · small run
SWE-bench Verified lab43.4% ✎▁▁▁▁▁▁▂▅█▆▁▁ beats 33/136medium · avg 47%▇▇ 4.1%at its level · small run
LiveCodeBench v6 (window unspecified) lab73.5% ✎▂▃▄▄▅██▆▄▇▄▆ beats 40/67—not counted yet: no other model in the same run to compare it with
SWE-bench Multilingual lab23.0% ✎▁··▃▃▁▄▇█▂▂▁ beats 1/63—not counted yet: no other model in the same run to compare it with
chat & writing█··········█ #2 of 2 * · 8.9% of its score · 1 test
testresultvs the others in its rundifficultycountswhy
IFEval (aggregation unspecified) lab96.7% ✎▁▂▄▂▂▄▆█▇▄▂▃ beats 142/143very easy · avg 97%▇▇▇▇ 8.9%maxed out · small run
web searchnot ranked on this task yet · 1 test
testresultvs the others in its rundifficultycountswhy
Humanity's Last Exam with tools lab13.9% ✎▃▃▁▂▃▃▃█▇▇▆▂ beats 0/58—not counted yet: no other model in the same run to compare it with
other tests5 not in a task · 18% of its score
testresultvs the others in its rundifficultycountswhy
Humanity's Last Exam lab7.2% ✎▁▁▁▄▅▆▆██▆▁▃ beats 8/109very hard · avg 7.0%▇▇▇▇ 8.9%too hard for it · small run
IFBench (aggregation unspecified) lab73.0% ✎▃▇▂▄▅▅▂▄▄▇█▇ beats 61/84easy · avg 74%▇▇▇ 8.0%at its level · small run
GPQA Diamond lab63.4% ✎▁▁▁▁▂▄▅▅█▇█▄ beats 122/318medium · avg 69%▇ 0.8%blunt · at its level · small run
RULER 128k lab65.6% ✎▃··▅▄█▅█▇·▃▃ beats 12/33—not counted yet: no other model in the same run to compare it with
AutomationBench (Public) lab41.3% ✎█····▄····▄▄ beats 4/4—not counted yet: no other model in the same run to compare it with

$ 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 LFM3 1.7B · needs ~1.04 GB · from huggingface.co/LiquidAI (official)
    chat in the terminal
    llama cli -hf LiquidAI/LFM3-1.7B-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf LiquidAI/LFM3-1.7B-GGUF:Q4_K_M

LM Studio

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

    download it (macOS / Windows / Linux) ↗

  2. Run LFM3 1.7B · needs ~1.04 GB · from huggingface.co/LiquidAI (official)

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/LiquidAI/LFM3-1.7B-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 LFM3 1.7B · needs ~1.04 GB · from huggingface.co/LiquidAI (official)
    chat in the terminal
    ollama run hf.co/LiquidAI/LFM3-1.7B-GGUF:Q4_K_M

Official weights and documentation: huggingface.co/LiquidAI ↗

$ how the score was worked out · Liquid AI own-model benchmark… 64%

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
Liquid AI own-model benchmark facts · lab-reported7Factual vendor-reported score (not licensed content)
prior · pull toward the middle
weighted average= 35.0

not counted (7)

Liquid AI own-model benchmark factsAutomationBench (Public) [correct]41.3%fewer than three models have this test yet, or this source ran it for this model alone
Liquid AI own-model benchmark factsMCP-Atlas (public set) [correct]32.6%fewer than three models have this test yet, or this source ran it for this model alone
Liquid AI own-model benchmark factsSWE-bench Multilingual [correct]23.0%fewer than three models have this test yet, or this source ran it for this model alone
Liquid AI own-model benchmark factsTerminal-Bench 2.1 [correct]12.9%fewer than three models have this test yet, or this source ran it for this model alone
Liquid AI own-model benchmark factsLiveCodeBench v6 (window unspecified) [correct]73.5%fewer than three models have this test yet, or this source ran it for this model alone
Liquid AI own-model benchmark factsHumanity's Last Exam with tools [correct]13.9%fewer than three models have this test yet, or this source ran it for this model alone
Liquid AI own-model benchmark factsRULER 128k [correct]65.6%fewer than three models have this test yet, or this source ran it for this model alone

$ 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