Claude Slashes Sonnet 5.5 Cache Reads, Unveils Cheaper Haiku 5.5

Claude's Aggressive Pricing & Performance Leap

Anthropic's latest announcements regarding Claude Sonnet 5.5 and the introduction of Claude Haiku 5.5 signal a strategic shift towards enhanced affordability and performance, particularly for high-volume, cost-sensitive applications. The halving of cache read prices for Sonnet 5.5 to $0.10 per million tokens is a substantial reduction that directly translates to a roughly 20% cost decrease for long-running workloads. This move is crucial for making advanced LLM deployments more economically viable for businesses grappling with the operational expenses of extensive AI integration. Furthermore, the launch of Claude Haiku 5.5 positions it as a leading contender in the small, fast, and capable model segment. Promising up to a 75% cost reduction compared to its predecessor, Haiku 4.5, and significant improvements in coding, computer use, and knowledge work, it's designed to excel at repetitive tasks like summarization and classification. Its availability on major cloud platforms (AWS, Google Cloud, Azure) further democratizes access and encourages broader adoption. The introduction of an adjustable effort setting for Haiku 5.5 is particularly noteworthy, offering developers granular control over the cost-intelligence trade-off, a feature highly sought after in optimizing AI resource utilization. This flexibility is especially beneficial for tasks under 100,000 tokens, which represent the majority of requests for this model tier.

Key Points

• Claude has halved the price of cache reads on Sonnet 5.5 to $0.10 per million tokens, reducing overall running costs by approximately 20% for long-running tasks.
• Claude Haiku 5.5 has been launched as the cheapest, fastest, and most capable small model from Anthropic, costing around 75% less than Haiku 4.5.
• Haiku 5.5 shows significant improvements in coding, computer use, and knowledge work, and is optimized for high-volume, cost-sensitive tasks.
• Haiku 5.5 features an adjustable effort setting, allowing users to balance cost and intelligence per task.
• The model is particularly cost-effective for tasks under 100,000 tokens, which constitute about 90% of requests to its predecessor.
• Haiku 5.5 is available on major cloud platforms: AWS, Google Cloud, and Microsoft Azure.

---

📖 Source: One more thing: we’re halving the price of cache reads on Claude Sonnet 5.5, to $0.10 per million to...