AI Benchmark Model
K2 Horizon 3.7B
Benchmark scores, pricing, speed, and model comparisons for K2 Horizon 3.7B.
Key scores
The K2 Horizon 3.7B is a foundation model developed by the MBZUAI Institute of Foundation Models, officially released on September 3, 2026. This model is characterized by an intelligence index of 16.2 and a coding index of 26.1. Regarding its performance on standardized evaluations, the model achieved a GPQA score of 0.692, an HLE score of 0.139, a SciCode score of 0.22, and an LCR score of 0.623333333333333. The provider has established a pricing structure of $0.00 per 1 million tokens for input, output, and blended usage. Specific metrics concerning output speed and time to first token remain unknown, and the model's performance on math-related benchmarks has not been disclosed.
Benchmark results
4 tracked benchmark rows for K2 Horizon 3.7B.
GPQA Reasoning
69.2% LCR Other
62.3% SciCode Other
22.0% HLE Other
13.9%