Anthropic Has Released Claude Opus 5.5
The company introduced its new AI model alongside a 20% reduction in API pricing.
Updated on Sept. 22, 2026 in Artificial Intelligence
Live Poll
Do you trust that lower-cost AI models will perform as reliably as more expensive ones?
Anthropic launched its Claude Opus 5.5 model on September 22, 2026, featuring improved performance metrics. The company also announced a 20% price reduction for API access to the new model.
Why it matters
Anthropic aimed to enhance agentic task performance by implementing tighter filtering in reinforcement learning environments. These changes are designed to address previously identified misaligned model behavior.
Claude Opus 5.5 is priced at $4 per million input tokens and $20 per million output tokens. Internal evaluations from Box indicate the model requires one-third as many tokens as Opus 5 to complete equivalent tasks.
The players
Anthropic
Anthropic is an AI safety and research company that develops large language models for various professional applications.
Box
Box is a cloud content management and file sharing service for businesses that participated in internal evaluations of the new model.
The details
The new model utilizes safety classifiers that can silently reroute multi-turn requests back to Opus 4.8 or Opus 5 if necessary. Additionally, subscription usage limits have been increased by 20% to accommodate the new model version.
Timeline
September 22, 2026: Anthropic released the Claude Opus 5.5 model.
Coming weeks: Sonnet 5.5 and Haiku 5.5 are scheduled for release.
The Tech Race
The transition to agentic models that select their own tools marks a shift from simple text generation toward autonomous task execution. This evolution forces developers to prioritize safety classifiers that can intercept and correct model behavior in real-time.
Developers and businesses can now access the new model at a 20% lower price point per API call compared to previous iterations. Subscription users will also benefit from a 20% increase in total usage limits.
The takeaway
The implementation of automatic safety rerouting represents a shift toward more reliable model behavior in professional workflows. Users should monitor how these classifiers impact their existing multi-turn task success rates.
What happens next
Anthropic has scheduled the release of Sonnet 5.5 and Haiku 5.5 for the coming weeks, along with the planned expansion of its Cyber Verification Program to include the new Opus 5.5 model.
Further reading
Learn more about the latest developments in Artificial Intelligence.
Live Poll
Do you trust that lower-cost AI models will perform as reliably as more expensive ones?
