Anthropic released Claude Haiku 5.5 on Oct. 7, cutting its price to match OpenAI's cheapest model. The new rate, $0.10 per million input tokens and $0.50 per million output tokens up to 100,000 tokens, exactly matches GPT-6 Luna at that tier, according to Anthropic and independent developer Simon Willison.
The previous Haiku, 4.5, launched almost a year ago priced at $1 per million input tokens and $5 per million output tokens, Anthropic said. On OSWorld 2.1, a benchmark for computer-use tasks, Haiku 5.5 scored 72.4%, up from Haiku 4.5's 15.7%, according to Anthropic.
Business customers reported gains in Anthropic's announcement. Asana said it saw "over a 30% reduction in latency for task completions," and Box said the new model "scored 11 points higher than Haiku 4.5 at about half the latency," the companies said, according to Anthropic.
The price match does not hold at every scale. Willison found Haiku 5.5 uses a less generous tokenizer than its predecessor, consuming about 25% more tokens for the same prompts, a hidden cost increase despite the matching headline rate. He also noted that beyond 100,000 tokens, Haiku's price jumps to $0.50 per million input and $2.50 per million output, while Luna's rises only to $0.20 and $0.75, making Luna cheaper for longer-context work.
Builders picking a model on price alone should check their actual token counts, not the sticker. The parity holds for short prompts, but a stingier tokenizer and a steeper price curve past 100,000 tokens mean a workload that looked evenly matched on paper can cost more with Haiku once it runs long.