Models

Every model we have tested, and what it did in each test. The same model can appear on two routes, so numbers are never mixed across routes.

ModelCompanyTestsAnswers
Claude Fable 5.1Anthropic4187
Claude Opus 5.5Anthropic4187
Claude Sonnet 5.5Anthropic4187
Claude Haiku 5.5Anthropic4237
GPT-6 AstraOpenAI249
GPT-6.1 SolOpenAI249
GPT-6 Luna (mini)OpenAI249
gpt-oss-120b (open)OpenAI249
Gemini 3.1 Pro PreviewGoogle249
Gemini 3.8 FlashGoogle249
Gemma 4 31B (open)Google249
Grok 4.7xAI249
Llama 4 MaverickMeta249
DeepSeek V4 ProDeepSeek249
DeepSeek V4.1 FlashDeepSeek249
Mistral Large 4Mistral249
Qwen3.8 Max PrimeAlibaba Qwen249
Kimi K3Moonshot249
GLM 5.3Z.ai249