Antropic·OpenAI launches new AI model with cost-effectiveness amid claims of speed control
Quick Look
- Antropic and OpenAI unveiled 'Claude Opus 5.5' and 'GPT-6 Sol·Luna' respectively, competing for cost-effective models that maintain or improve performance and lower costs.
- This is analyzed as a strategy to defend market share and increase sales ahead of the IPO while insisting on speed control.
AI-generated summary
Why It Matters
Antropic and OpenAI had previously warned of AI risks and insisted on controlling the speed of development, but this time they began competing in the market by launching new models that maintained or improved performance and lowered costs.
(San Francisco = Yonhap News) Correspondent Kwon Young-jeon = Antropic and OpenAI, which warned of the risks of artificial intelligence (AI) and raised the theory of controlling the speed of development, side by side introduced new AI models that emphasize price efficiency.
Antropic unveiled its new AI model ‘Claude Opus 5.5’ on the 22nd (local time).
This is a model that delivers performance at the same level as the company's top model, 'Claude Fable 5.1', but at a lower cost.
Antropic said of this model, "This is the first model we have introduced since we called for speed control of cutting-edge AI," and added, "It performs at the level of Fable 5.1 in most tasks and costs 40% less than the previous Opus 5."
In particular, it received a better evaluation than the top model, recording higher scores than Fable 5.1 in three coding-related performance indicators (benchmarks) such as 'Terminal Bench', 'Frontier Code', and 'Cursor Bench' and knowledge work-related performance indicators.
In particular, Antropic introduced that Opus 5.5 recorded the highest score among Claude models released so far as a result of testing the 'Alignment' section, which allows AI to set goals similar to human intentions, in 2,000 scenarios.
In addition, it was explained that safety measures related to sensitive fields such as biology and cybersecurity, and 'distillation' used for copying AI models, were increased to a similar level as Fable 5.1 to reduce risk factors.
On the same day, OpenAI also competed in the 'cost-effectiveness' competition by unveiling 'GPT-6 Sol' and 'GPT-6 Luna', lower models of the top model 'GPT-6 Astra'.
OpenAI introduced, “If GPT-6 Astra introduced a new generation of intelligence, Sol and Luna help spread the benefits of intelligence by broadening the horizon of cost efficiency.”
OpenAI began a more aggressive price competition, offering the application program interface (API) fee for GPT-6 Sol·Luna at half the price of its predecessor, GPT-5.6 Sol·Luna.
Artificial Analysis, an AI performance evaluation company, set Claude Opus 5.5's intelligence index (AAII) at 58 points, higher than Fable 5.1, placing it in first place. GPT-6 Sol earned 48 points and tied for 5th place.
The fact that both companies are competing to launch new models with cost-effectiveness is interpreted as a strategy to maintain market dominance while promoting 'speed control theory'.
The strategy is to prevent customers from leaving to other latecomers or China's open source models by replacing the top models 'Claude Mythos Fable' and 'GPT Astra' by increasing the performance of the next top model to a level comparable to the top model and lowering the price.
Both companies are likely to calculate that in order to increase sales as they are about to go public (IPO), they must provide a model with high 'cost-effectiveness' to the market.
What to Watch
AI outlook — possibilities, not facts
The launch of Antropic and Open AI's cost-effective models will lead to pressure on other AI companies to lower prices.
Likely · Within months
The IPOs of both companies are likely to take place within the next six months.
Possible · Within months
Open Questions
- When will the new model actually be commercialized?
- How will the competitive landscape with the Chinese open source model develop?
- What is the timing and size of the IPOs of the two companies?
