Name: LFM2.5-2.6B
Author: Liquid AI

LFM2.5-2.6B

Liquid AI

Liquid AI's hybrid model for on-device agentic workloads. Its card identifies it as a pure reasoning model that always thinks before answering and supports 131,072-token context.

Benchmark scores

Benchmark	Score
AA-LCR v1.1	5.7
AA-Omniscience Index	-10.9
Artificial Analysis Intelligence Index v4.3	8.4
Artificial Analysis Intelligence Index v4.3.2	8.4
Artificial Analysis Omniscience Accuracy	4.3
Artificial Analysis Omniscience Hallucination Rate	16.0
Artificial Analysis Openness Index	38.9
Capability	113.1
CritPt (avg@5)	0.0
GDPval-AA v2	202
GDPval-AA v2.1	-22.2
GPQA Diamond (Artificial Analysis avg@5)	55.8
Humanity's Last Exam (Text-Only)	6.2
SciCode 1.0.1 (Artificial Analysis)	14.4
Terminal-Bench 2.1	4.5
𝜏³-Banking	7.2