Google’s Gemini 3.0 beats GPT-5 on coding benchmarks (it’s complicated)
Gemini 3.0 posts a higher SWE-bench Verified than GPT-5 in Google’s own report — but the gap shrinks when you control for tool budget.
Gemini 3.0 posts a higher SWE-bench Verified than GPT-5 in Google’s own report — but the gap shrinks when you control for tool budget.
Headlines say Gemini 3.0 > GPT-5 on coding. Reality: the benchmark allowed Gemini double the tool calls in the published comparison, and on apples-to-apples runs the two are within noise.
We’ll re-run with a fixed tool budget once API access opens up.
Get tomorrow's best AI project in your email. With a guide that works. Free. No spam.