Anthropic’s newest small AI model costs as little as 10 cents per million input tokens, making it one of the company’s cheapest options for developers.
The company has launched Claude Haiku 5.5, a faster, lower-cost model designed for high-volume tasks such as summarization, classification and AI agent workflows. Anthropic says the model costs about 75% less to run on average than Haiku 4.5 while delivering substantial performance improvements.
For requests up to 100,000 tokens, Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens, down from $1 and $5 for Haiku 4.5. Above that threshold, prices rise to $0.50 and $2.50 per million tokens, still 50% below the previous model.
The pricing puts Haiku 5.5 on par with OpenAI’s GPT-6 Luna at the lower tier, although token prices alone do not determine which model is cheaper for a completed task. Token consumption, retries and accuracy can all change the final bill.
Haiku 5.5 makes gains in Anthropic’s benchmarks
Anthropic’s testing shows Haiku 5.5 making large gains over Haiku 4.5. It scored 72.4% on the OSWorld 2.1 offline subset, compared with 15.7% for its predecessor, and 39.2% on Terminal-Bench 4.0 versus 0% for Haiku 4.5.
It also scored higher than GPT-6 Luna on several of Anthropic’s listed evaluations, including GDPval-AA v2.1 and OSWorld. Sonnet 5.5 remained ahead on the benchmarks supplied by Anthropic.
Haiku 5.5 is also the first Haiku model with adjustable effort settings, letting developers decide how much computing power to use for a task.
Why cheaper models matter for AI agents
Anthropic is not pitching Haiku 5.5 as a replacement for its larger models. Instead, it wants the model to take over the smaller jobs inside larger AI workflows. That could make a difference for companies running agents at scale.
A Sonnet or Opus model can handle the main assignment while Haiku 5.5 performs repetitive subtasks that previously were too expensive to run continuously. Early customer feedback points in that direction.
Asana’s Aaron Vinh said, “Compared with the model we use today, we saw over a 30% reduction in latency for task completions and up to 2.5x faster inference per agent turn.”
Those results reflect one customer’s experience and may not apply to every workload.
More must-read AI coverage
- SS&C Intralinks DealCentre AI vs. Datasite: Which platform is built for the future of dealmaking?
- SS&C Intralinks FundCentre AI vs. Juniper Square: Which platform better supports modern private markets fund managers?
- Why Data, Not Models, Determines AI Success
- The Rise of the AI-Native Factory: How Physical AI Is Transforming Manufacturing
Anthropic is cutting more than Haiku prices
The company also cut Sonnet 5.5 cache-read prices by 50%, from $0.20 to $0.10 per million tokens. Anthropic estimates that will reduce Sonnet’s cost on most agentic workloads by about 20%.
Max and Team subscribers will also receive monthly API credits: $100 for Max 5x, $200 for Max 20x and up to $500 pooled for Team accounts. The credits can be used across Anthropic’s models.
For developers, Haiku 5.5 could make high-volume automation more affordable. Customer-support routing, document summaries, database lookups and agent subtasks are among the workloads that could benefit.
However, the lowest pricing tier applies only to requests within the 100,000-token threshold, while larger models such as Sonnet and Opus may remain better suited to complex coding and reasoning tasks.
The opportunity for developers is to use Haiku 5.5 for simpler, repetitive work while reserving more capable models for demanding assignments. That approach could help businesses reduce AI costs without sacrificing performance where it matters most.
Want to learn more AI tips, tricks, and prompting techniques? TechRepublic readers get free 7-day access to The Neuron Academy, our practical learning platform designed to help professionals use AI more confidently at work. Browse all lessons →
Let us teach you How to Talk to AI for free! Try our six-minute course at The Neuron Academy and learn a few simple ways to write better prompts and get more useful results from AI, or browse our other AI course for free for seven days. Check out all the lessons here →