Name: K2 Horizon 0.9B
Author: LLM360

K2 Horizon 0.9B

LLM360

The compact dense K2 Horizon reasoning model from IFM's K2/LLM360 program. The released main checkpoint uses multi-domain on-policy distillation and 128K context.

Benchmark scores

Benchmark	Score
AA-Briefcase Elo	0.0
AA-Briefcase v1.1 Elo	0.0
AA-LCR v1.1	6.3
AA-Omniscience Index	-72.8
Artificial Analysis Intelligence Index v4.3	3.0
Artificial Analysis Intelligence Index v4.3.2	3.0
Artificial Analysis Omniscience Accuracy	6.9
Artificial Analysis Omniscience Hallucination Rate	85.7
AutomationBench-AA 1.0.6	0.2
Capability	114.3
CritPt (avg@5)	0.0
GDP.pdf (Artificial Analysis, All-pass)	0.8
GDP.pdf (Artificial Analysis, Mean Criteria)	26.4
GDPval-AA v2	-174
GDPval-AA v2.1	-426
GPQA Diamond (Artificial Analysis avg@5)	29.3
Humanity's Last Exam (Text-Only)	5.4
SciCode 1.0.1 (Artificial Analysis)	6.9
Terminal-Bench 2.1	1.1
Terminal-Bench 4.0 (Artificial Analysis)	0.0
𝜏³-Banking	4.9