Anthropic: Claude 3 Haiku

Claude 3 Haiku is Anthropic's text-and-image model with a 200,000-token context window and a 4,096-token output ceiling. It supports tool use, making it suitable for agentic workflows, but it does not offer native reasoning or confirmed structured output support. The large context window allows it to process lengthy documents or multi-image inputs in a single pass. At $0.25 per million input tokens and $1.25 per million output tokens, it sits toward the affordable end of the market, which is its clearest argument for consideration. Benchmark coverage is thin, with a blended score of 18.2 across only two benchmarks, so its measured capability remains largely unproven relative to better-tested alternatives. Teams running high-volume, cost-sensitive tasks that involve long context or image inputs may find it worth shortlisting, but buyers who need confidence from broad benchmark data should treat those numbers with caution.