LFM2.5 24B-A2B

+ contribute

LFM2.5 24B-A2B is an open-weights language model from Liquid AI, released October 2026. Its published size is 24B parameters (2.45B active per token) and a 131k-token context window. Licence: lfm1.0.

Model information sources

Release date ↗

No comparable score yet. Its published results are below, but they don't overlap enough with other models to calculate a score.

$ how it compares

Scored LLMs released closest to LFM2.5 24B-A2B, out of 1188 with a score, then more from Liquid AI.

#	model	score	where	RAM	released
5	Muse Wonder Full Meta	82±2	Muse Code only		2026-10
15	DeepSeek V4.1 Flash 1017 DeepSeek	79±1	api $0.26/M		2026-10
16	Muse Wonder Meta	79±1	api $8.00/M		2026-10
28	Claude Haiku 5.5 Anthropic	76±1	api $0.20/M		2026-10
32	MiniMax M3.1 Flash MiniMax	75±2	api $0.26/M		2026-10
	more from Liquid AI
521	LFM2.5 2.6B Liquid AI	53±1	local	~2.3GB	2026-07
534	LFM2.5 8B-A1B Liquid AI	53±1	local	~5.5GB	2026-05
566	LFM2 24B-A2B Liquid AI	51±1	local	~14GB	2026-02

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasksnot ranked on this task yet · 4 tests

test	result	vs the others in its run	counts	why
SWE-bench Verified lab	51.0% ✎	▁▁▁▁▁▁▂▅█▆▁▁ beats 42/134	—	not counted yet: no other model in the same run to compare it with
BFCL V4 (December 2025 archive) lab	65.0% ✎	▁▁▁▂▇▆▆▆█▂▄▂ beats 97/112	—	not counted yet: no other model in the same run to compare it with
τ³-bench Banking lab	8.0% ✎	▂▂·▅█▂▃▃▃▆▂▅ beats 11/24	—	not counted yet: no other model in the same run to compare it with
WebArena-Verified lab	26.8% ✎	█······█·█·█ beats 0/3	—	not counted yet: no other model in the same run to compare it with

codingnot ranked on this task yet · 2 tests

test	result	vs the others in its run	difficulty	counts	why
SWE-bench Verified lab	51.0% ✎	▁▁▁▁▁▁▂▅█▆▁▁ beats 42/134		—	not counted yet: no other model in the same run to compare it with
SciCode (level unspecified) lab	32.7% ✎	▁····▃▄▄▅█▇▃ beats 14/37		—	not counted yet: no other model in the same run to compare it with

chat & writingnot ranked on this task yet · 1 test

test	result	vs the others in its run	difficulty	counts	why
IFEval (aggregation unspecified) lab	96.8% ✎	▁▂▄▂▂▄▆█▇▄▂▃ beats 142/142		—	not counted yet: no other model in the same run to compare it with

mathnot ranked on this task yet · 1 test

test	result	vs the others in its run	difficulty	counts	why
MATH-500 lab	96.0% ✎	▁▃▂▃▄▇▅▇█▄▂▁ beats 60/85		—	not counted yet: no other model in the same run to compare it with

other tests3 not in a task · 0.0% of its score

test	result	vs the others in its run	counts	why
GPQA Diamond lab	74.4% ✎	▁▁▁▁▂▄▅▅█▇█▄ beats 170/315	—	not counted yet: no other model in the same run to compare it with
Humanity's Last Exam lab	7.4% ✎	▁▁▁▄▅▆▆██▆▁▃ beats 9/108	—	not counted yet: no other model in the same run to compare it with
IFBench (aggregation unspecified) lab	74.7% ✎	▃▇▂▄▅▅▂▄▄▆█▇ beats 63/83	—	not counted yet: no other model in the same run to compare it with

$ where to run it

llama.cpp · recommended

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 ↗
Run LFM2.5 24B-A2B · needs ~14.4 GB · from huggingface.co/LiquidAI (official)
chat in the terminal
llama cli -hf LiquidAI/LFM2.5-24B-A2B-GGUF
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf LiquidAI/LFM2.5-24B-A2B-GGUF

LM Studio

Install LM Studio · Open the app once before using the lms command.
download it (macOS / Windows / Linux) ↗
Run LFM2.5 24B-A2B · needs ~14.4 GB · from huggingface.co/LiquidAI (official)
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/LiquidAI/LFM2.5-24B-A2B-GGUF -y
Then run lms ls to see its name, and lms chat followed by that name

Ollama

Install Ollama
download it (macOS / Windows) ↗
Linux
curl -fsSL https://ollama.com/install.sh | sh
Run LFM2.5 24B-A2B · needs ~14.4 GB · from huggingface.co/LiquidAI (official)
chat in the terminal
ollama run hf.co/LiquidAI/LFM2.5-24B-A2B-GGUF

Official weights and documentation: huggingface.co/LiquidAI ↗

$ published results · 10 results, not enough overlap for a score

test	result	source and conditions
BFCL V4 (December 2025 archive)	65.0%	Liquid AI own-model benchmark facts · lab-reported Individual Liquid own purple-column figures read from pinned benchmark image; vLLM evaluation, precision not specified. SWE Verified uses OpenCode; IFEval aggregation not stated. Benchmark-specific sampling settings in card. No competitor columns or AA metrics.
τ³-bench Banking	8.0%	Liquid AI own-model benchmark facts · lab-reported Individual Liquid own purple-column figures read from pinned benchmark image; vLLM evaluation, precision not specified. SWE Verified uses OpenCode; IFEval aggregation not stated. Benchmark-specific sampling settings in card. No competitor columns or AA metrics.
WebArena-Verified	26.8%	Liquid AI own-model benchmark facts · lab-reported Individual Liquid own purple-column figures read from pinned benchmark image; vLLM evaluation, precision not specified. SWE Verified uses OpenCode; IFEval aggregation not stated. Benchmark-specific sampling settings in card. No competitor columns or AA metrics.
GPQA Diamond	74.4%	Liquid AI own-model benchmark facts · lab-reported Individual Liquid own purple-column figures read from pinned benchmark image; vLLM evaluation, precision not specified. SWE Verified uses OpenCode; IFEval aggregation not stated. Benchmark-specific sampling settings in card. No competitor columns or AA metrics.
Humanity's Last Exam	7.4%	Liquid AI own-model benchmark facts · lab-reported Individual Liquid own purple-column figures read from pinned benchmark image; vLLM evaluation, precision not specified. SWE Verified uses OpenCode; IFEval aggregation not stated. Benchmark-specific sampling settings in card. No competitor columns or AA metrics.
IFEval (aggregation unspecified)	96.8%	Liquid AI own-model benchmark facts · lab-reported Individual Liquid own purple-column figures read from pinned benchmark image; vLLM evaluation, precision not specified. SWE Verified uses OpenCode; IFEval aggregation not stated. Benchmark-specific sampling settings in card. No competitor columns or AA metrics.
IFBench (aggregation unspecified)	74.7%	Liquid AI own-model benchmark facts · lab-reported Individual Liquid own purple-column figures read from pinned benchmark image; vLLM evaluation, precision not specified. SWE Verified uses OpenCode; IFEval aggregation not stated. Benchmark-specific sampling settings in card. No competitor columns or AA metrics.
MATH-500	96.0%	Liquid AI own-model benchmark facts · lab-reported Individual Liquid own purple-column figures read from pinned benchmark image; vLLM evaluation, precision not specified. SWE Verified uses OpenCode; IFEval aggregation not stated. Benchmark-specific sampling settings in card. No competitor columns or AA metrics.
SWE-bench Verified	51.0%	Liquid AI own-model benchmark facts · lab-reported Individual Liquid own purple-column figures read from pinned benchmark image; vLLM evaluation, precision not specified. SWE Verified uses OpenCode; IFEval aggregation not stated. Benchmark-specific sampling settings in card. No competitor columns or AA metrics.
SciCode (level unspecified)	32.7%	Liquid AI own-model benchmark facts · lab-reported Individual Liquid own purple-column figures read from pinned benchmark image; vLLM evaluation, precision not specified. SWE Verified uses OpenCode; IFEval aggregation not stated. Benchmark-specific sampling settings in card. No competitor columns or AA metrics.

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit 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 2026-10-21 08:35 UTC · sources and licences · json