On October 7, Anthropic released Claude Haiku 5.5, described as its "fastest, cheapest, and most capable small model yet." Engineered for high-volume, cost-sensitive workloads—summaries, context compaction, database queries, and classification—it cuts inference pricing for under-100K-token requests to 0.10 p er mi l l i o nin p u tt o k e n s an d 0.50 per million output tokens, slashing average running costs by roughly 75%. The move signals Anthropic's push to extend its tiered pricing model lineup and defend market share in an increasingly competitive generative AI ecosystem. The company also updated its 2026 Usage Policy on October 8, reinforcing governance commitments. Separately, DeepSeek open-sourced infrastructure components for Huawei's Ascend compute platform, strengthening domestic AI hardware-software synergy and signaling China's push for self-reliant AI compute stacks.
Sources： threerouter.com
Top comments (0)