Name: K2 Horizon 3.7B
Author: LLM360

K2 Horizon 3.7B

LLM360

The small dense K2 Horizon model from IFM's K2/LLM360 program, with a 524,288-token context and request-level reasoning effort controls.

Benchmark	Score
AA-Briefcase Elo	793
AA-Briefcase v1.1 Elo	788
AA-LCR v1.1	62.3
AA-Omniscience Index	-28.6
Artificial Analysis Intelligence Index v4.3	16.2
Artificial Analysis Intelligence Index v4.3.2	15.6
Artificial Analysis Omniscience Accuracy	10.8
Artificial Analysis Omniscience Hallucination Rate	44.1
AutomationBench-AA 1.0.6	14.7
Capability	140.7
CritPt (avg@5)	0.0
GDP.pdf (Artificial Analysis, All-pass)	3.2
GDP.pdf (Artificial Analysis, Mean Criteria)	45.1
GDPval-AA v2	962
GDPval-AA v2.1	865
GPQA Diamond (Artificial Analysis avg@5)	69.2
Humanity's Last Exam (Text-Only)	13.9
SciCode 1.0.1 (Artificial Analysis)	22.0
Terminal-Bench 2.1	28.1
Terminal-Bench 4.0 (Artificial Analysis)	0.0
𝜏³-Banking	19.6