Ctrl K
Back to models

Liquid AI

LFM2.5-2.6B

  • Released
  • Open weights
  • LFM Open License v1.0
Compare
Intelligence Index

Reasoning setting

8.4
Estimated · Artificial Analysis v4.3.2
API price from
$0.00 in / $0.00 out / 1M tokOpenRouter · Liquid · FP8
Context window
131.07K tok

Performance profile

ReasoningRank #92 of 107
55.8%
GPQA Diamond
CodingRank #93 of 94
14.4%
SciCode

Capabilities

Context window

Modalities

Measured setting

Integration

Reasoning trade-offs

Source (opens in a new tab)
Intelligence Index v4.3.2
ReasoningEstimated
Intelligence Index
8.4

Pricing & access

API route
OpenRouter · Liquid · FP8API
Tier & contextStandard

USD per 1M tokens

Input
$0.00
Output
$0.00

Run locally

GGUF
Context
Estimated memory

~3.7 GiB

RAM + VRAM

Estimate assumptions

One text chat at the selected context, with a 16-bit cache and app allowance. Operating system and image processing are extra; optimized caches can use less.

8 GiB RAMwithout GPU offload

Memory budgetOne text chat
Weights
1.6 GiB
Context + state
0.1 GiB16-bit cache
App allowance
2 GiB
Download
1.7 GB
Disk space
~7 GB
Available software
Ollama (opens in a new tab)LM Studio (opens in a new tab)Jan (opens in a new tab)KoboldCpp (opens in a new tab)llama.cpp (opens in a new tab)
Get GGUF (opens in a new tab)
Set up in Ollama · Q4_K_M · 8K context

Save the following as Modelfile. The context matches the estimate above.

FROM hf.co/lmstudio-community/LFM2.5-2.6B-GGUF:Q4_K_M
PARAMETER num_ctx 8192

Then run these commands in your terminal:

ollama create llmcurrent-local -f Modelfile
ollama run llmcurrent-local

Ollama import instructions (opens in a new tab) · For LM Studio or Jan, open the app’s model search, enter lmstudio-community/LFM2.5-2.6B-GGUF, choose Q4_K_M, and set context to 8,192 tokens in the app’s model settings. Software links above open each app’s setup guide.

Transformers setup (opens in a new tab)

Benchmarks

Full leaderboardSource (opens in a new tab)

Humanity’s Last Exam

Frontier reasoning

LFM2.5-2.6B · Reasoning#103 of 126

  1. 1Claude Opus 5.5Max61.4%
  2. 2Claude Fable 5.1Max59.1%
  3. 3Gemini 4 ArgonHigh57.1%
  5. 100Qwen3 Omni 30B A3B (Reasoning)Reasoning7.5%
  6. 101Celeris-1Standard6.8%
  7. 102Granite 4.2 3BReasoning6.6%
  8. 103LFM2.5-2.6BReasoning6.2%
  9. 104Llama 3.2 Instruct 11B (Vision)Standard5.5%
  10. 105NVIDIA Nemotron Nano 12B v2 VLReasoning5.5%
  11. 106Ministral 3 3BStandard5.4%

Peers use their best measured result on this benchmark.