Anthropic slashes Haiku 5.5 pricing, cuts Sonnet cache costs

Anthropic has released Claude Haiku 5.5, pricing the new small model at roughly a quarter of Haiku 4.5’s cost, while halving cache read prices for Claude Sonnet 5.5. The updates follow Opus 5.5’s launch on September 22, continuing a push to reduce inference costs across its model lineup.

The move extends a trend Anthropic has pursued since late September, when it shipped Sonnet 5.5 with a 30% speed improvement and lower per-token pricing. At the time, both Anthropic and OpenAI were adjusting prices to address enterprise concerns over per-token billing, which a recent QumulusAI report identified as a barrier for agentic workloads. OpenAI’s GPT-6.1 Sol, announced the day after Sonnet 5.5, similarly targeted cost-sensitive users with a lower-priced alternative to its delayed Astra model.

Haiku 5.5’s pricing suggests Anthropic is now competing more aggressively in the small-model segment, where speed and affordability often matter more than raw capability. The cache price cut for Sonnet 5.5 could also help retain customers running repetitive or high-volume tasks, where cached prompts reduce token consumption. While neither change alters the models’ underlying performance, the cumulative effect may shift adoption patterns among startups and enterprises balancing cost and output quality.

The timing raises questions about whether OpenAI or other providers will respond with further price adjustments. For now, Anthropic’s strategy appears to prioritize cost efficiency over feature expansion, a calculus that could reshape how enterprises evaluate model tiers.

Sources: siliconangle.com