BREAKING
BJD, BJP spar over move to set up nuclear power plantIs the superstar taking a gamble?Father and son killin in accidentGedela Raju: Raghu Kunche wins accoladesRoad safety drive lacks progressShravan Reddy speaks about his multi-faceted role in ‘Agadha’Collectors asked to ensure minimum school staffingRakul mesmerizes in leopard-print dressLow pressure again, IMD forecasts heavy rains‘Shiva Gadi Zindagi’ first-single unveiled
Technology

Claude Opus 5 is here: Anthropic claims top coding benchmarks at lower price than Fable 5

hellohimachalse@gmail.com 25 Jul 2026 25 Jul 2026

New Delhi: Anthropic has introduced Claude Opus 5, its latest artificial intelligence model designed for coding, scientific research, and complex knowledge work. The new model offers the performance of the company’s flagship Claude Fable 5, but at half the price, the company says, it’s a more practical option for everyday use. Claude Opus 5 has become the standard model for Claude Max customers and is the most robust model offered on the Claude Pro plan.

As AI firms race one another for top performance, price and efficiency, the launch arrives. Anthropic says the new Claude Opus 5 outperforms its predecessor, Claude Opus 4.8, across multiple software engineering and reasoning benchmarks while maintaining the same pricing of $5 per million input tokens and $25 per million output tokens.

Claude Opus 5 focuses on coding, reasoning and research

Claude Opus 5 matches or outperforms state-of-the-art models on various coding and knowledge-work benchmarks, such as Frontier-Bench and GDPval-AA. The model also excels on software engineering benchmarks like CursorBench and AutomationBench, the company said, offering better performance per dollar than competing models in several evaluations.

The company also emphasised enhancements in scientific studies. Anthropic indicates that Claude Opus 5 is better than Opus 4.8 in structural biology, organic chemistry and bioinformatics tasks. It also produces vivid visual output and is better able to check its own work when creating answers.

Built to complete complex tasks with fewer errors

Anthropic presented some internal and initial customer examples to illustrate the capabilities of the model. One example of this was when Claude Opus 5 was asked to reconstruct a machine part in 3D without being able to directly access the machine’s original drawing, according to the report. In another instance the model led to the discovery of the source of a software bug that was not being fully addressed by a community patch.

The company also announced that an engineer from a trading firm was able to create a market data feed for a new exchange within one session using Claude Opus 5. Anthropic attributed the model’s planning, checking and repeating to achieve successful tasks.

Safety remains a priority despite higher performance

The Claude Opus 5 model is the most aligned with the company’s AI safety principles to date, with less deceptive behaviour and higher adherence to the principles than prior Claude models, Anthropic said. The model, however, does not push the envelope on high-risk cybersecurity or biology skills, the company said, and it is still lagging the Mythos 5 model in offensive cyber work.

To minimise misuse, Anthropic has incorporated security measures to prevent penetration testing, generation of exploits, and other aspects of cybersecurity workflows. The company has also added automatic API fallbacks that enable a requested API to be rerouted to another API that fits the safety requirements, rather than rejected.

Claude Opus 5: Pricing and availability

Claude Opus 5 has been launched today on Anthropic’s platforms Claude API. It will have the same pricing as Claude Opus 4.8, at $5 per million input tokens and $25 per million output tokens. Anthropic has also unveiled an extra fast mode that is around 2.5 times faster than regular for a double price and new developer capabilities like mid-conversation tool changes and automatic API fallbacks.

Related News