Claude Haiku 4.5: benchmarks, price and rank

Its headline results

The benchmarks that count most in the rating, with its place among the models measured the same way, and where each number came from.

• Vending-Bench 2$459

  The model runs a simulated vending machine business for a year on its own: finding suppliers, ordering stock, setting prices and paying the daily fees.

  55th of 60. Independent result from Epoch AI.

• SimpleQA Verified13.2%

  Short factual questions with one checked answer, asked without web search.

  75th of 78. Independent result from Epoch AI.

• LMArena text1,397

  People put a question to two anonymous models and vote for the better answer.

  85th of 107. Independent result from LMArena.

• LMArena WebDev1,330

  People ask two anonymous models to build the same web page or app and vote for the better one.

  87th of 98. Independent result from LMArena.

• Finance Agent v231.0%

  Real financial analysis work: finding figures in company filings and documents, multi-step calculations and written analysis.

  21st of 23. The maker's own number, from Anthropic, via LLM Stats.

• SWE-bench Verified73.3%

  Real bug reports and feature requests from popular open-source Python projects. The model has to change the code so the project's hidden tests pass.

  22nd of 56. The maker's own number, from Anthropic, via LLM Stats.

• GPQA Diamond71.2%

  Graduate-level questions in biology, physics and chemistry, written by PhDs so that searching the web does not give the answer away.

  100th of 103. Independent result from Epoch AI.

• ARC-AGI-24.0%

  Visual puzzles of colored grids, each with a new rule to work out from a few examples.

  62nd of 69. Independent result from Epoch AI.

• Epoch Capabilities Index142.8

  One score built from dozens of benchmarks, set on a common scale so that models tested on different benchmarks can still be compared.

  85th of 94. Independent result from Epoch AI.