Anthropic released Claude Opus 5 on July 24, 2026, positioning it as a cost-efficient alternative to its most powerful model, Fable. Opus 5 delivers performance close to Fable on many tasks at half the price—$5 per million input tokens and $25 per million output tokens (same as the previous Opus 4.8).

The model is now available across all Anthropic platforms: Claude.ai, the API, and enterprise deployments. Anthropic claims Opus 5 excels at software engineering, knowledge work, and coding tasks—benchmarking competitively on Frontier-Bench and GDPval-AA against models that cost double.

What’s Different in Opus 5

The headline feature is effort control. Users can now toggle how hard Claude works on a task—low, medium, or high effort. This lets developers dial in latency and quality tradeoffs on a per-request basis. A chatbot might use low effort for quick responses. A code review tool might use high effort to catch edge cases. It’s a knob previously unavailable in Claude’s pricing model.

Performance gains are incremental, not transformative. Opus 5 doesn’t match Fable on reasoning-heavy benchmarks. But for the 80% of real-world tasks that don’t require maximum capability, the speed-and-cost advantage makes Opus 5 the obvious choice.

The Market Signal

This is Anthropic’s answer to a crowded market. OpenAI has GPT-4 Mini. Meta has Llama variants. Mistral has optimized lightweight models. Google has Gemini 1.5 Flash. Every frontier lab is now racing to the middle—competing on bang-for-buck rather than pure capability. That’s good news for developers building real products, less good for labs trying to justify premium pricing on Fable or comparable models.

The effort control feature is interesting because it signals how Anthropic thinks about capability. Rather than releasing discrete models at discrete price points, the company is treating capability as a slider. Deployment teams can now run Claude at three speeds on one model, reducing infrastructure fragmentation and simplifying decision-making.

Who Benefits

Software engineers building internal tools, customer-facing chatbots, and automation pipelines. Teams already on Claude’s API will likely upgrade to Opus 5 and see immediate cost savings. New teams evaluating Claude against GPT-4 Mini now have a harder decision—Opus 5 is more capable, same price per token. Anthropic’s bet: capability wins.