Anthropic’s new Claude 4.1 dominates coding tests days before GPT-5 arrives

AI company Anthropic has upgraded its flagship model, Claude Opus 4.1, improving its performance in software engineering tasks.
It achieved a score of 74.5% on SWE-bench Verified, a benchmarking test, higher than OpenAI’s o3 model (69.1%) and Google’s Gemini 2.5 Pro (67.2%).
While this positions Anthropic as a leader in AI-powered coding assistance, nearly half of its $3.1 bn A P I re v e n u e d e p e n d so n tw oc u s t o m ers, C u rsor an d M i croso f t^{'} s G i t H u b C o p i l o t, w hi c h t o g e t h er g e n er a t e$ 1.4bn.
The new model has been released ahead of the anticipated launch of OpenAI’s GPT-5, heightening fears that Anthropic is trying to maintain its dominance in the AI coding market.
Safety concerns remain paramount, with Opus 4.1 launched with strict safety protocols after previous models exhibited concerning behaviours during AI blackmail tests.

Fast Feed