Claude and Gemini are now better than ChatGPT for most tasks

With Opus 4.5 and Gemini 3 Pro, ChatGPT is now clearly behind. On raw intelligence and speed, both models beat GPT-5 Thinking on many benchmarks. More importantly, they lead on my vibe checks and those of most others.¹

For real world use, and especially for coding, Opus 4.5 is a step change.

The Claude web app has many features that Gemini lacks (e.g. skills, memory, attach Google Doc by pasting a URL). So: Opus 4.5 should now be your default model.² But Gemini sometimes wins on harder tasks, so use both.

Gemini sometimes wins on harder tasks, and tasks that require web search. ChatGPT remains best for web search. Use all three.

Some details:

• ChatGPT remains best for web search, followed by Gemini. Claude is a distant third, but may be improving. (Confidence: High)
• Claude still doesn't use web search proactively enough, so I see it hallucinate more often than Gemini and ChatGPT. (Confidence: High)
• For many simple tasks, you'll get answers of roughly the same quality from all models. (Confidence: High)
• ChatGPT is better at French than the other models. (Confidence: Medium)
• Claude Code dominates Codex CLI at the moment—complete reversal from October! (Confidence: High)