unknown

Claude Opus 4.5 demonstrates superior capabilities for multi-agent orchestration and long-running engineering tasks compared to Gemini 3. The model excels at delegating work to sub-agents, managing browser automation workflows, and building full-stack applications through agentic workflows. With pricing reduced by one-third from the previous version, Opus 4.5 offers state-of-the-art performance at 60 tokens per second. The model's strength lies in prompt engineering for sub-agents and executing complex, multi-step development workflows including planning, building, hosting, and automated testing through browser agents operating in isolated sandboxes.

How Claude Opus 4.5 DESTROYED Gemini 3 on Launch Day

Bobby Iliev

What do you mean “how”?
Assumably Anthropic has better/smarter people or work in a different way than Google engineers because every model they release is better than any Gemini version.
We cannot really know because nobody sat with both teams, for long enough time, to be able to compare why one team keeps being better than the other.

Ah yes, the classic ‘make a 3% bump look like a moon landing’ chart design

sonnet 4.5 destroyed gemini3 pro and it came out way before lmao

You have to donate your kidney if you want to work with Opus 4.5.