Anthropic cut Claude Sonnet 5.5 cache-read pricing from $0.20 to $0.10 per million tokens on 7 October and began rolling out monthly API credits of up to $500 for Team subscribers, The Decoder reported.
Key points
- Anthropic says the 50% cut in Sonnet 5.5 cache-read pricing should reduce costs for most agentic tasks by about 20%.
- Max-5x subscribers get $100 in monthly API credits, Max-20x subscribers get $200, and Team subscribers receive up to $500.
- The changes accompanied the release of Claude Haiku 5.5, which has separate prices for shorter and longer requests.
Sonnet 5.5 cache reads cost $0.10
The reduction applies to cache reads, one part of the Sonnet 5.5 API bill. Its listed rates are $2.00 per million input tokens and $10.00 per million output tokens, while cache writes cost $2.50 per million tokens. Anthropic expects the cheaper reads to lower costs for most agentic tasks by about 20%, according to The Decoder.
That projected saving concerns tasks rather than every token a customer buys. The size of a customer’s bill depends on which of the listed charges its workload incurs. Anthropic’s estimate is therefore a different measure from the 50% cut to the cache-read rate.
AI Affairs covered the Sonnet 5.5 release and its faster output before the caching change. The model’s listed input and output prices remain $2.00 and $10.00 per million tokens, respectively.
Max and Team subscribers receive credits
The monthly credits vary by subscription. Max-5x subscribers get $100 for API use, and Max-20x subscribers get $200. Team subscribers receive up to $500 per month. The Decoder reported that users can spend the credits through the API while experimenting with tools, applications and agents.
The credits and the cache-read cut change different parts of a customer’s cost. The new read price applies to Sonnet 5.5 caching, while the credits give eligible subscribers a monthly amount to spend through the API. Both changes accompanied Anthropic’s release of Claude Haiku 5.5.
Anthropic is also updating its Python and TypeScript software development kits with beta support for computer use and browser use. Those updates sit alongside the pricing changes rather than altering Sonnet 5.5’s listed input and output rates.
Haiku 5.5 has two price tiers
Haiku 5.5 starts at $0.10 per million input tokens and $0.50 per million output tokens for requests below 100,000 tokens, VentureBeat reported. Above 100,000 tokens, the respective rates are $0.50 and $2.50 per million. Cache reads cost $0.01 per million tokens in the lower tier and $0.05 in the higher tier.
Anthropic estimates Haiku 5.5 costs about 75% below Haiku 4.5 on average, factoring in request length and token use. Its prices for shorter requests are 90% below Haiku 4.5’s listed token rates, compared with a 50% reduction for longer requests. Anthropic says approximately 90% of Haiku 4.5 requests were in the shorter category.
The lower-tier Haiku input and output rates match OpenAI’s GPT-6 Luna base rates. VentureBeat reported that Luna’s higher pricing begins above 272,000 input tokens, a different threshold from Haiku’s 100,000-token boundary. Its pricing comparison also lists OpenAI’s GPT-6.1 Sol at $2.00 per million input tokens and $10.00 per million output tokens, the same listed rates as Sonnet 5.5.
Anthropic says Haiku 5.5 uses an updated tokenizer that consumes somewhat more tokens for equivalent work than Haiku 4.5. The company’s estimated 75% workload saving accounts for that change as well as the mix of shorter and longer requests.
