Anthropic Announces New Flagship Model "Claude Opus 5"

Anthropic has announced its new flagship AI model "Claude Opus 5". The model achieves high scores in coding and knowledge work fields while significantly reducing token pricing compared to competing models.

In the AI industry, "cost efficiency" has emerged as an important evaluation axis alongside model performance competition. While high-performance models tend to incur higher usage costs, Anthropic has positioned the balance between performance and price as a key selling point. This movement reflects the industry-wide awareness that price often becomes a bottleneck when companies integrate AI into actual business operations.

According to the announcement, Claude Opus 5's token pricing is set at approximately half that of "Fable 5", the competing model considered equivalent to OpenAI's "GPT-4.5". In terms of performance, Anthropic claims the model achieved top-tier scores on coding and knowledge work benchmarks. Additionally, on "ARC-AGI-3", a new benchmark for measuring problem-solving ability, it achieved a score of 30.2%, equivalent to approximately 4 times the level of "GPT-5.6 Sol".

ARC-AGI-3 is a benchmark that measures how well AI can solve novel problems without relying on existing learned patterns. It is positioned as testing reasoning and flexibility in application rather than simple knowledge memorization or reproduction. This score difference can be viewed as indicating ability differences beyond mere processing speed or knowledge quantity. However, it should be noted that whether benchmark results directly translate to superiority in actual business use depends on the application and environment.

The significance of this announcement extends beyond performance alone. The fact that Anthropic simultaneously presented price competitiveness against the traditional "high performance equals high price" formula is noteworthy from the perspective of expanding options in the enterprise AI market. Particularly for organizations that have cautiously proceeded with AI adoption due to cost concerns, this pricing strategy could serve as a consideration trigger.

Meanwhile, the performance figures presented by Anthropic are based on the company's own announcement, and verification by independent third parties has not yet been confirmed. Since benchmark comparisons can sometimes be conducted by each company under favorable conditions, evaluation under actual usage scenarios is needed. As verification through real-world use by developers and enterprises progresses, the actual capabilities of Claude Opus 5 are expected to become clearer.

AI model competition is expanding not only in pursuit of peak performance but also in the direction of "whether sufficient performance can be provided at an affordable price". Anthropic's current move can be seen as one case accelerating this trend. As model commoditization (generalization and cost reduction) progresses across the industry, which differentiation strategy each company presents will become the focal point of future competition.