By 2026, AI agents will run workflows — but only if we stop chasing 'super agents' and design them to stay in their lanes.
Anthropic claims that Claude Sonnet 4.5 scored 77.2 percent on the SWE bench benchmark, beating GPT-5 and Gemini 2.5 Pro.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback