ARTIFICE DAILY AI + CYBER SECURITY BRIEFING

Artificial Intelligence

Claude Haiku 5.5 launches with lower API costs

Anthropic announced Claude Haiku 5.5 as a cheaper and faster small model, while also lowering Sonnet 5.5 cache-read prices and adding monthly API credits for paid subscribers.

AnthropicAI

Anthropic introduced Claude Haiku 5.5 and described it as its lowest-cost, quickest and most capable small model so far. It said the model was aimed at high-volume, cost-sensitive work such as summaries, compaction, database queries and classification, and that it could also be used alongside Opus 5.5 and Claude Sonnet 5.5 on coding tasks.

The company said Haiku 5.5 cost about 75% less to run than Haiku 4.5 on average. It said the reduction was 90% for requests up to 100,000 tokens and 50% above that, after taking tokenizer changes into account.

Anthropic also said it reduced Claude Sonnet 5.5 cache-read pricing to $0.10 per million tokens from $0.20. It said that made most agentic work about 20% cheaper and accompanied a monthly API credit for Max and Team subscribers.

Max 5x subscribers were to get $100 a month, Max 20x subscribers $200, and Team subscribers up to $500 shared across users. Anthropic said Haiku 5.5 was available on Amazon Web Services, Google Cloud, Microsoft Azure and the Claude Platform, and that its Python and TypeScript SDKs were being updated to support computer use and browser use in beta.

Named in this story

Organisations

Amazon Web Services
One of the platforms carrying Haiku 5.5
Google Cloud
One of the platforms carrying Haiku 5.5
Microsoft Azure
One of the platforms carrying Haiku 5.5

Products and systems

Claude Haiku 5.5
Anthropic’s new small model
Claude Sonnet 5.5
Its cache-read price was cut in half
Claude Haiku 4.5
The earlier Haiku model used for price comparisons
Opus 5.5
A larger model paired with Haiku 5.5 on coding work
Claude Platform
The platform where developers could start with claude-haiku-5-5
Claude Python and TypeScript SDKs
Were updated to add beta support for computer use and browser use
Max 5x
A subscriber tier that received monthly API credits
Max 20x
A subscriber tier that received monthly API credits
Team
A subscriber tier that received pooled monthly API credits

How the source tells it

The piece was promotional launch copy, using superlatives, self-reported performance claims and repeated new-product framing.

  • hype superlative description of the new model at launch
  • vendor boosterism company-run benchmark charts and early customer feedback used to support the claims
  • novelty repeated launch and rollout language around the model, pricing, credits and SDK updates
  • reassurance the safety section presented alignment gains and tighter safeguards as improvements