Gemini and Claude Outpace ChatGPT in AI Benchmarks

What happened

Gemini and Anthropic’s Claude are now outperforming ChatGPT in key AI benchmarks. Gemini 3 has achieved higher scores than OpenAI’s GPT-5 Pro on major reasoning and visual comprehension tests. Anthropic’s Claude continues to lead in safety and consistency, while Gemini excels in speed and cost efficiency. These shifts mark a significant change in the competitive landscape for leading AI models.

Why it matters

• Gemini and Claude’s advances challenge ChatGPT’s dominance in AI performance and reliability.
• Higher benchmark scores and lower costs for Gemini make it a strong contender for enterprise and research use.
• Anthropic’s focus on safety and consistency positions Claude as a preferred option for sensitive or regulated tasks.

Key facts

1. 1. Gemini 3 scored almost twice as high as GPT-5 Pro on the ARC AGI 2 reasoning test.
2. 2. Gemini 3 Pro holds the top spot for visual comprehension, surpassing a 1500 score on the text leaderboard.
3. 3. Gemini 3 achieved its benchmark results at one-tenth the cost per task compared to GPT-5 Pro.
4. 4. Anthropic’s Claude remains the safest and most consistent model, often refusing to assist with potentially harmful requests.

Key Voices

• The Verge noted that Gemini 3’s benchmark scores and cost efficiency mark a major shift in AI model leadership.
• Anthropic emphasized Claude’s safety and reliability, positioning it as the preferred model for sensitive applications.