Anthropic has released Claude Haiku 5.5, a smaller Claude model aimed at high-volume and speed-sensitive tasks, while sharply reducing its API pricing. The company is charging $0.10 per million input tokens and $0.50 per million output tokens for prompts of up to 100,000 tokens, compared with $1 and $5 for the previous Haiku 4.5 model.
The launch puts Haiku 5.5 at the same headline price as OpenAI’s GPT-6 Luna, which launched on September 22. Anthropic also says the new model improves on its predecessor in several tests of computer use, coding and professional work. However, the model remains positioned below the company’s larger Sonnet 5.5 model in at least one coding benchmark.
Lower costs target high-volume AI workloads
Claude Haiku 5.5 is designed for jobs where organizations need quick responses across large numbers of requests. Anthropic identified document summarization, database queries, customer-support conversations and browser operation as intended uses. These workloads generally involve repetitive or relatively straightforward tasks rather than highly creative work.
AI providers bill developers according to tokens, the small pieces of text that models process. Anthropic’s new prices apply to prompts containing up to 100,000 tokens. For longer prompts, the company has cut pricing by 50% compared with Haiku 4.5.
Haiku 4.5 cost $1 per million input tokens and $5 per million output tokens. That means the new standard rates represent a 90% reduction. Anthropic estimates the average saving at about 75%, partly because roughly 90% of requests to the older model were below the 100,000-token threshold and because Haiku 5.5 divides text into slightly more tokens.
The lower price is significant for developers embedding AI into customer-facing services, where even small per-request costs can accumulate at scale. Anthropic has not described the launch as a replacement for its larger models; instead, Haiku 5.5 is the faster, lower-cost option in the Claude lineup.
Benchmark results show stronger performance
According to Decrypt, Haiku 5.5 scored 72.4% on the offline subset of OSWorld 2.1. The benchmark tests whether an AI system can operate a real computer through lengthy, multi-step tasks and awards partial credit. Haiku 5.5’s result was above the 48.9% recorded by OpenAI’s GPT-6 Luna.
The new model also performed better than Luna on Terminal-Bench 4.0, which gives AI agents professional tasks to complete by entering commands independently. Haiku 5.5 completed 39.2% of the tasks correctly on the first attempt, compared with 16.4% for Luna and 0% for Haiku 4.5.
Anthropic’s larger Claude Sonnet 5.5 scored 70.6% on that coding-focused test, showing that Haiku 5.5’s lower cost does not make it the strongest performer across every evaluation. On GDPval-AA v2.1, which uses an Elo-style rating to assess professional work across 44 occupations, Haiku 5.5 received a 1,620 rating. Luna scored 1,437, while Haiku 4.5 scored 735.
Haiku 5.5 is also the first Haiku model to include an adjustable effort setting. The control lets users trade speed and cost for potentially more capable responses. At the same time, Anthropic said it had reduced Sonnet 5.5’s cache-read price to $0.10 per million tokens. Cache reads refer to text the model has already processed and can use again at a discounted rate.
Benchmark results do not guarantee that every response will be accurate. A simple logic test conducted during the reporting produced a rapid but incorrect answer, underscoring the need to check outputs rather than trust them automatically.
Availability, credits and market picture
Haiku 5.5 is available through the Claude website, Amazon Web Services, Google Cloud and Microsoft Azure under the model name “claude-haiku-5-5.” The release completes Anthropic’s promised Claude 5.5 model series after Opus 5.5 on September 22 and Sonnet 5.5 on September 28. Anthropic released Haiku 5.5 on Wednesday.
Anthropic is also introducing monthly API credits this week. Max 5x subscribers will receive $100 in credits, Max 20x subscribers will receive $200, and Team plans can receive up to $500 pooled across users.
Moneyiar market data at 09:19 on October 8, 2026, showed mixed performance among the listed digital assets:
- ApeCoin (APE): $0.1479586863, down 0.61% over 24 hours and 1.48% over seven days.
- Decimal (DEL): $0.3857105613, up 0.17% over 24 hours but down 10.73% over seven days.
- RedStone (RED): $0.1517873648, up 0.63% over 24 hours and down 16.94% over seven days.
These market figures are separate from Anthropic’s product pricing and do not indicate how investors should respond to the launch.
