Claude Opus 5.5 scores 66 on the Artificial Analysis Coding Agent Index

At max effort in Claude Code, Claude Opus 5.5 scores 66 on the Artificial Analysis Coding Agent Index, the highest score Artificial Analysis has measured. That is 6 points above Opus 5 at 60 and 4 points above Claude Fable 5.1 at 62. The index gives equal weight to Terminal-Bench 4.0, DeepSWE v1.1, and SWE-Atlas-QnA.

Opus 5.5 reaches 63.1% on Terminal-Bench 4.0, up from 54.5% for Opus 5; 68.4% on DeepSWE v1.1, up from 62.5%; and 66.4% on SWE-Atlas-QnA, up from 62.1%. The largest increase is 8.6 percentage points on Terminal-Bench.

Artificial Analysis says Anthropic cut Opus pricing to $4 per million input tokens and $20 per million output tokens, from $5 and $25 for Opus 5, and cut cache reads to $0.20 per million from $0.50. Opus 5.5’s cost per task is still $13.04, 21% above Opus 5’s $10.79, because it uses about 15.6 million tokens per task versus 11.4 million, including about 2.4 times as many output tokens. No lower-cost model in the comparison matches its score.