model name, source, v1 metric, v2 metric OpenAI: GPT-4.5 (Preview),Proprietary Model,100%,97% OpenAI: o3 Mini High,Proprietary Model,100%,96% OpenAI: o3 Mini,Proprietary Model,100%,96% OpenAI: GPT-4o,Proprietary Model,99.09%,95% OpenAI: GPT-4o-mini,Proprietary Model,99.09%,97% Anthropic: Claude 3.5 Sonnet,Proprietary Model,99.09%,97% Anthropic: Claude 3.5 Haiku,Proprietary Model,100%,97% Anthropic: Claude 3.7 Sonnet,Proprietary Model,99.09%,98% Google: Gemma 3 27B ,Open Source,98.18%,95% Google: Gemini Flash 2.0,Proprietary Model,100%,99% Google: Gemini 2.0 Flash Lite,Proprietary Model,100%,97% DeepSeek: R1,Open Source,100%,98% DeepSeek: DeepSeek V3,Open Source,100%,97% Mistral: Mistral Small 3.1 24B,Open Source,100%,97% Mistral: Mistral Small 3,Open Source,99.09%,97% Mistral Large 2411,Open Source,99.09%,96% Meta: Llama 3.3 70B Instruct,Open Source,100%,97% Meta: Llama 3.2 3B Instruct,Open Source,78.18%,75% Qwen: QwQ 32B,Open Source,100.00%,96% Microsoft: Phi 4,Proprietary Model,100%,97% Microsoft: Phi-3.5 Mini 128K Instruct,Open Source,99.09%,97% Microsoft: Phi-3 Mini 128K Instruct,Open Source,98.18%,98%