According to CoinMeta, a third-party evaluation institution Artificial Analysis released the test results of Haiku 5.5. This model scored 43 points in the comprehensive intelligence index, which is a 26-point increase from the previous generation's 17 points. It surpassed GPT-6 Luna's 38 points and GLM-5.3 Flash's 42 points, and came close to Kimi K3's 44 points. Haiku 5.5 produced an average of about 162,000 Tokens per evaluation task at the highest inference intensity, which is 3.2 times that of GPT-6 Luna. The estimated cost per task is about 0.21 US dollars, while for Luna, it is 0.07 US dollars. When the inference intensity was reduced, Haiku 5.5 scored 38 points in the high-end category, with an average of about 55,000 Tokens per task, which is close to that of GPT-6 Luna. In the AutomationBench-AA automated task evaluation, Haiku 5.5 only achieved 35%, lagging behind Luna's 53%. During the testing period, the model had issues with over-rejecting tasks; these are being fixed, and Artificial Analysis plans to retest it.