Anthropic announced Claude Haiku 5.5 on October 7, aiming at fast, inexpensive work such as coding subagents, summaries, and repetitive tasks. The company describes it as its fastest model at standard speed, with stronger results than GPT-6 Luna on several reported evaluations.
Low prices for prompts up to 100,000 tokens
The following standard Claude API prices are in US dollars per 1 million tokens:
| Token category | Prompt up to 100,000 tokens | Prompt over 100,000 tokens |
|---|---|---|
| Input | $0.10 | $0.50 |
| Output | $0.50 | $2.50 |
| Cache reads | $0.01 | $0.05 |
The official pricing documentation applies the higher rates to a request whose total prompt exceeds 100,000 tokens. Prompt length includes cache reads and cache writes. Each request is priced separately, so long conversation histories can move a request into the more expensive tier.
Benchmark gains and the right workload
Anthropic reports higher scores than GPT-6 Luna on Terminal-Bench 4.0, knowledge-work evaluations, and its computer-use test. These are the company's benchmark results; the useful question for a developer is how reliably the model completes their own small, frequent tasks.
The launch reinforces how competitive inexpensive models have become alongside the flagship race. Read the Anthropic announcement for the evaluations, or our Claude Opus 5.5 overview for the larger model in the family.
Adapted from the original Chinese article, published on October 8, 2026 (China Standard Time).
Comments NOTHING