Performance Alone Is Not Enough: 'Value-for-Money' Competition Heats Up, Even in Frontier AI

[Financial News] The center of gravity in the global artificial intelligence (AI) competition is shifting from performance to cost-effectiveness. Global AI developers are focusing on lowering the prices of their flagship models or improving operating efficiency so they can perform more tasks at the same cost.
As new AI models that put price competitiveness front and center continue to be launched despite calls to slow the pace of AI development, the battle to lead AI adoption is expected to intensify.
According to industry sources on the 4th, global AI developers are stepping up their price competition by introducing a series of cost-effective models, alongside improving AI model performance. These models focus on token pricing, processing speed, cache costs and the actual cost of completing tasks.

Google's flagship AI model, Gemini 4 Argon, was launched at promotional prices of $2 per million input tokens and $10 per million output tokens. That is just half the price of Anthropic's comparable flagship model, Claude Opus 5.5, which costs $4 per million input tokens and $20 per million output tokens.
Anthropic's Claude Opus 5.5 delivers performance comparable to that of the flagship Claude Fable 5.1 across most tasks, while reducing operating costs by 40% from its predecessor, Opus 5. Its application programming interface (API) pricing is $4 per million input tokens and $20 per million output tokens. Compared with its predecessor, both input-token and output-token prices are 20% lower. Cache-read pricing, which is important for agentic and coding tasks, has been cut by 60%. The mid-tier Claude Sonnet 5.5 is 30% faster than its predecessor and requires fewer tokens per task, reducing actual costs by up to 30%.
OpenAI also put price-performance at the forefront with the launch of GPT-5.6 last July. It cut the price of Luna, its fastest, lightest low-cost model, by 80%, while reducing the price of Terra, its balanced model, by 20%. However, it kept the price of Sol, its highest-performance model, unchanged.

Global AI developers are focusing on price competitiveness in what is seen as an effort to target companies that remain cautious about adopting AI models because of the return on their investment. For enterprise customers, the cost of using a model to complete a single task is becoming more important than its absolute performance. As AI agents repeatedly call models to perform tasks such as coding, research and document writing, the burden of token fees is rising rapidly. In other words, the AI development paradigm is shifting. The focus is no longer only on developing the highest-performing AI models, but also on making sufficiently capable AI available at the lowest possible cost.
In response, AI companies are segmenting their models by price and performance. Alongside cutting prices for their highest-performing models, they are improving the capabilities of mid-tier models and launching separate low-cost, high-speed models.
An industry official said, "The more heavily a company uses AI, the much greater the perceived impact will be than the actual price reduction for the model. The strategy is to win new corporate customers by lowering the cost of AI model services offered to businesses and accelerate the creation of an AI ecosystem centered on the company's own services."
[email protected] Jang Min-kwon Reporter