Science & Tech · AI · 2 hrs ago
AI competition shifts from model power to cost and productivity
AI competition is shifting toward cost and productivity, not just model performance or how widely AI is used.
OpenAI and Anthropic have introduced models that cost less or offer better performance for the same price.
Businesses are matching tasks to models, using lighter, cheaper systems for simple work and more powerful ones for complex analysis.
Some AI providers are also charging for results, such as documents produced or customer inquiries resolved, instead of usage alone.
Samsung and SK Telecom are changing work practices in South Korea to make AI part of how teams operate, rather than only a tool for individual employees.
Samsung uses a mix of on-device and cloud AI, while SK Telecom is developing AI agents with work roles and permissions.
Companies may first see productivity fall as staff adjust and receive training, but SK Telecom's CEO says it could rise sharply after that transition.
AI competition is shifting from model performance and scale toward cost efficiency and productivity, according to Yonhap.
OpenAI and Anthropic have introduced models with lower prices or improved performance.
Companies are choosing models according to task difficulty and increasingly considering charges based on results rather than usage.
Samsung and SK Telecom are focusing on changes to work practices and organisational systems as well as AI adoption.
SK Telecom is developing a system in which AI agents have employee identities, roles and permissions.
- Who
- AI model companies including OpenAI and Anthropic, and businesses including Samsung and SK Telecom.
- What
- Competition is shifting toward cost efficiency and productivity, with companies also redesigning work around AI.
- When
- The report was published on 29 October 2026. It also describes initiatives launched in March and events in September.
- Where
- The report discusses global AI companies and South Korean businesses.
- Why
- Rising AI use makes the cost of computing and token consumption important, while companies seek greater productivity from AI.
This story does not have two clearly opposing sides.
SK Telecom began its “one AI agent per person” project, encouraging all employees to build and use AI agents for work.
Samsung stressed at the Samsung AI Forum that AI development should shift its focus from scaling up to efficiency.
SK Telecom CEO Jeong Jae-heon described an “AX J-curve” in a lecture at Seoul National University.
Yonhap reported that AI competition was shifting toward cost efficiency and productivity.
- OpenAI model
- GPT-6.1 Sol offers performance close to its top model at one-fifth the price.
- Anthropic model
- Claude Opus 5.5 has a 20% lower token price.
- AWS Bedrock
- Offers more than 100 AI models.
- Finn
- Charges $0.99 per customer inquiry resolved by an AI support agent.
- SK Telecom project
- One AI agent per person” began in March 2026.
1 source
What people are saying
Companies that establish token-efficiency systems in the initial J-curve phase and turn token consumption into real business value will take the lead in the AI era.