GPT-5.6 Has a New Rival From Anthropic: Here’s What Claude Opus 5 Brings

Anthropic has launched Claude Opus 5, its latest artificial intelligence model for coding, professional work and complex agent-based tasks. The release gives OpenAI’s GPT-5.6 another competitor as AI companies increasingly focus on balancing performance with operating costs.

Claude Opus 5 sits below Fable 5 in Anthropic’s model lineup but is designed as a more practical option for regular professional use. Anthropic claims the model approaches Fable 5 performance on several tasks while costing half as much through the API.

The model is available across Claude’s paid subscription plans, Claude Code, Claude Cowork and the Claude API.

Improved Coding and Knowledge Work

Anthropic says Claude Opus 5 provides a major performance improvement over Opus 4.8,

particularly for software engineering, research, data analysis and multi-step professional tasks.

On the company’s Frontier-Bench evaluation, Opus 5 reportedly achieved more than twice the performance of Opus 4.8 at a lower cost per completed task. At maximum effort on CursorBench 3.2, it performed within 0.5 percent of Fable 5’s highest score while costing half as much per task.

These figures are based largely on Anthropic’s testing and evaluations involving early-access partners. Independent testing will be needed to determine how consistently the model competes with GPT-5.6 in real-world workloads.

Adjustable Effort Levels

Claude Opus 5 allows developers to control how much computational effort the model applies to a request. Lower effort settings can reduce response time and token consumption, while higher levels are intended for difficult coding, reasoning and research tasks.

The available options include low, medium, high, xhigh and max. This gives developers more control over the balance between capability, speed and cost when building applications.

Opus 5 also supports a one-million-token context window and can generate outputs containing up to 128,000 tokens.

New Tools for AI Agents

Anthropic has introduced mid-conversation tool changes in beta. Developers can add, remove or replace tools available to Claude during a conversation without invalidating the existing prompt cache.

Another beta feature, automatic fallbacks, can route requests blocked by Opus 5’s safety systems to another suitable model instead of returning an error. Both additions are intended to make production AI agents more flexible and reliable.

Price and Availability

Claude Opus 5 costs $5 per million input tokens and $25 per million output tokens, matching Opus 4.8. Fable 5 costs twice as much at $10 and $50, respectively.

A Fast mode runs approximately 2.5 times faster but costs twice the standard Opus 5 API rate. The model is now the default option for Claude Max subscribers and the strongest model available on Claude Pro.