MoonshotAI: Kimi K2.7 Code

MoonshotAI: Kimi K2.7 Code, a coding-focused model with a 262144 context length, demonstrates moderate speed performance, ranking in the 27th percentile across benchmarks. Its pricing is also moderate, placing it in the 26th percentile. A standout feature is its exceptional reliability, achieving a 97% success rate, indicating minimal technical failures and consistent evaluable responses. In terms of specific performance, the model exhibits strong capabilities in hallucination avoidance (96.0% accuracy) and reasoning (94.0% accuracy, 76th percentile), suggesting a robust understanding of factual boundaries and complex problem-solving. Its core strength in coding is evident with 91.0% accuracy, placing it in the 65th percentile for this critical category. Mathematics also shows solid performance at 92.0% accuracy. However, the model shows weaknesses in instruction following, with a 42.0% accuracy (31st percentile), and general knowledge (93.0% accuracy, 31st percentile), indicating areas for improvement in handling complex multi-step directives and broad factual recall. Ethics performance is also relatively low at 93.0% accuracy (21st percentile). While its cost for mathematics is notably low (11th percentile), its overall speed and price are average.