Claude Haiku 5.5 Launches with 75% Price Reduction and Benchmark Improvements
- Published
- Oct 7, 2026 — 18:49 UTC
On October 7, 2026, Anthropic released Claude Haiku 5.5, which is priced approximately 75% lower than its predecessor, Haiku 4.5. The token pricing for Haiku 5.5 is set at $0.01 per million tokens for prompts up to 100,000 tokens and $0.05 per million tokens for prompts over that limit, compared to Haiku 4.5's $0.10 and $0.50 per million tokens, respectively. This represents a price drop of up to 90% for high-volume tasks.
Haiku 5.5 achieved a Knowledge Benchmark Score of 1,620 on GDPval-AA v2.1, significantly outperforming Haiku 4.5's score of 735. Additionally, it scored 45.9% on Humanity's Last Exam without tools and 57.4% with tools, while its Computer Use Score reached 72.4% on OSWorld-2.1. Notably, Haiku 5.5 also demonstrated a substantial improvement in Agentic Coding, scoring 39.2% on Terminal-Bench 4.0, a stark contrast to Haiku 4.5's 0.0%.
The release includes Sonnet 5.5, which features cache read costs of $0.10 per million tokens and cache writes at $2.50 per million tokens. Monthly API credits are available in tiers: Max-5x at $100, Max-20x at $200, and Team up to $500.
Anthropic indicated that Haiku 5.5 is tailored for cost-sensitive applications such as summarization, database queries, classification, and live customer support. This strategic pricing adjustment appears to be a direct response to competitive pressures from OpenAI's new GPT-6.1 series, as noted by Anthropic. While the lower prices suggest potential savings, the company cautioned that real-world savings may be smaller than indicated by the per-token rates.
By Turing Wire Newsdesk · Oct 7, 2026 · How we work →
Summarised from The Decoder's original report by the Turing Wire Newsdesk. Read the original for the full story.
Source: The Decoder
