Anthropic Launches Claude Opus 5 With Near-Fable Performance at Half the Cost

The new Claude Opus 5 model targets enterprise coding and knowledge work. It maintains lower pricing and lighter safety restrictions than Anthropic’s flagship Fable 5.

Anthropic has released Claude Opus 5, its latest large language model, positioning it as a lower-cost alternative to the company’s flagship Claude Fable 5. However, it retains much of its performance on enterprise workloads. According to Anthropic, Opus 5 delivers capabilities close to Fable 5 at roughly half the API price. It targets software development, office productivity, scientific research, and knowledge-intensive applications. The company announced the release on July 24 through its official channels. The new model is now available across Claude platforms and APIs.

Claude Opus 5 Focuses on Efficiency Rather Than Maximum Capability

Anthropic says Claude Opus 5 is intended for organizations that need strong reasoning and coding performance. These organizations do not want to pay premium pricing reserved for its frontier-class Fable 5 model.

According to the company, Opus 5 is priced at $5 per million input tokens and $25 per million output tokens. This matches the pricing of its predecessor, Opus 4.8. By comparison, Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens. Therefore, Opus 5 is available at half the inference cost.

Anthropic describes Opus 5 as “thoughtful and proactive,” emphasizing improvements in software engineering, knowledge work, and business automation. Yet, it remains below Fable 5 in the most demanding long-duration autonomous reasoning tasks.

Enterprise AI Competition Increasingly Centers on Cost

The launch comes during an increasingly competitive phase of the frontier AI market. Currently, major developers are racing to improve both performance and operating efficiency rather than focusing solely on benchmark leadership.

Anthropic product executive Dianne Penn told Reuters that the company’s objective is to make frontier intelligence more broadly accessible while maintaining rapid release cycles. Additionally, Reuters reported that users handling everyday office work or programming tasks are expected to benefit most from Opus 5. Meanwhile, Fable 5 remains better suited to complex projects requiring extended autonomous execution.

Industry observers note that enterprises are becoming more sensitive to AI inference costs as production deployments expand. Lower operating costs can significantly reduce expenses for organizations processing millions or billions of tokens every day.

Anthropic Also Highlights Updated Safety Characteristics

The company says Opus 5 incorporates updated safety improvements compared with previous generations.

According to Anthropic, internal testing found that Opus 5 is less likely than earlier Claude models to be manipulated into harmful behavior. At the same time, it requires fewer restrictive safeguards than Claude Fable 5. Similarly, Reuters reported that the reduced restrictions reflect the model’s lower capability for advanced cybersecurity exploitation compared with Fable 5.

This distinction follows Anthropic’s June launch of Claude Fable 5, which included additional safety controls because of its advanced capabilities. Anthropic previously explained that some sensitive requests submitted to Fable 5 could instead be routed to lower-capability models. This rerouting happens through automated safety systems.

Why The Launch Matters

Anthropic’s latest release illustrates a broader shift in the AI industry.

Rather than simply introducing increasingly larger models, developers are now emphasizing performance-per-dollar as enterprise customers seek predictable operating costs. Lower-priced frontier models may encourage wider deployment across software development, customer support, research, and business automation.

The announcement also arrives as AI developers face growing competition from both proprietary and open-weight models. During interviews surrounding the launch, Anthropic executives questioned whether emerging open-weight systems can consistently deliver comparable results on complex real-world projects. Nonetheless, lower-cost alternatives continue gaining attention.

For enterprise buyers, pricing differences may influence deployment strategies. This may allow premium models such as Fable 5 to handle the most demanding reasoning workloads. Meanwhile, lower-cost models like Opus 5 process routine production tasks.

Risks and Remaining Questions

Although Anthropic has published benchmark data supporting Opus 5’s performance, those evaluations largely originate from company testing. Independent benchmarking organizations and enterprise customers will likely provide additional evidence over the coming weeks. Their data will clarify real-world coding, reasoning, and productivity performance.

Another open question concerns adoption. Enterprise AI customers increasingly evaluate models based not only on benchmark scores but also on infrastructure costs, latency, safety behavior, and integration with existing software platforms.

What Happens Next

Anthropic has made Claude Opus 5 available through its Claude applications and API. Here, it replaces earlier Opus models while maintaining existing pricing for developers.

Market observers will now monitor enterprise adoption, third-party benchmark results, and customer migration from Opus 4.8. They will also look to see whether competing AI providers respond with pricing reductions or new model releases. Competition among frontier AI companies continues to intensify.