Claude Opus 5.5

Claude Opus 5.5 is a model from Anthropic, released Sep 22, 2026. What it could let Clay do, as a hypothesis: could trim the cost and the wait on long research runs, without expecting better answers than the older model gave. At least one benchmark was measured by a third party.

Claude Opus 5.5: 10 numbers, nine independent and one vendor-reported

• Artificial Analysis Intelligence Index

  58

  Compared to

  median: 26

  Source

  Artificial Analysis

  Verification

  Independent

• Output tokens per second

  97 tokens per second

  Compared to

  average (87)

  Source

  Artificial Analysis

  Verification

  Independent

• Cost per Intelligence Index task

  $5.98

  Compared to

  Not stated

  Source

  Artificial Analysis

  Verification

  Independent

• API price per 1M tokens

  $4.00 per 1M input tokens

  Compared to

  expensive, median: $2.00

  Source

  Artificial Analysis

  Verification

  Independent

• LMArena Top 10 Agents, Best Overall

  13.82 %

  Compared to

  Anthropic Claude Fable 5.1 (Max) 14.31 %

  Source

  LMArena

  Verification

  Independent

• LMArena Pareto Optimal Models, cost per task

  $1.57/task

  Compared to

  Anthropic Claude Fable 5.1 (Max) $5.38/task

  Source

  LMArena

  Verification

  Independent

• Cost to run versus Opus 5, as stated by the maker

  40% less

  Compared to

  Opus 5

  Source

  Anthropic Newsroom Sep 22, 2026

  Verification

  Vendor-reported

• Artificial Analysis Intelligence Index rank

  # 1 / 224

  Compared to

  224 models in this class

  Source

  Artificial Analysis

  Verification

  Independent

• Output speed (output tokens per second)

  97 tokens per second

  Compared to

  average (87)

  Source

  Artificial Analysis

  Verification

  Independent

• LMArena Pareto Optimal Models, Best Overall

  $1.57/task

  Compared to

  Anthropic Claude Fable 5.1 (Max) $5.39/task 14.31 %

  Source

  LMArena

  Verification

  Independent