Anthropic has launched Claude Haiku 5.5, a lightweight model designed for high-concurrency, low-latency, and cost-sensitive tasks. It succeeds the Haiku 4.5 model released nearly a year ago. The new model supports a context window of up to 1 million tokens and a maximum output length of 128,000 tokens.
The pricing structure for Haiku 5.5 features a tiered system. For requests under 100,000 tokens, the input cost is $0.10 per million tokens, and output is $0.50 per million tokens. If usage exceeds the 100,000-token threshold, prices increase fivefold to $0.50 for input and $2.50 for output per million tokens.
While the base price per million tokens is significantly lower than its predecessor, Haiku 5.5 utilizes a new tokenizer similar to the Sonnet 5.5 and Opus 5.5 models. This change results in a 25% to 30% increase in token consumption for identical tasks, leading to an estimated 75% reduction in actual operating costs rather than the 90% suggested by base pricing.
Anthropic has also reduced the cache read price for Claude Sonnet 5.5 by 50%, bringing it to $0.10 per million tokens. This adjustment is expected to lower the overall operating costs for agent-based workloads by approximately 20%.
The model is currently available via the Claude platform, Amazon Web Services, Google Cloud, Microsoft Foundry, and Claude Code, as well as through mobile and web applications. It is positioned to serve as a sub-agent for complex tasks handled by larger models like Opus 5.5 and Sonnet 5.5.
