Physix Frontier badgePhysix Frontier
PublishChinese | EN

Physix Frontier · News Briefing Card (TMTPost · Oct 11, 2026)

OpenAI and Anthropic Launch Small Models on Same Day, Big Tech Pivots to Good-Enough AI

KEY FACTS

  • On October 6, OpenAI published mathematical research completed by an internal frontier model, with each result requiring roughly three hours of ChatGPT Pro compute on average.
  • On October 7, Anthropic released the small model Haiku 5.5, suited for repetitive work such as summarization, classification, and database queries.
  • The same day, OpenAI pushed GPT-6 to more ChatGPT users, with free users getting the lighter Luna model.
  • Alibaba, Tencent, and ByteDance respectively integrated Qwen Office, WorkBuddy, and Doubao Work into office scenarios to handle specific tasks.
  • Alibaba's capital expenditure last quarter was 67.678 billion yuan, and Tencent's was 52.8 billion yuan, up 75% and 176% year-over-year respectively.

KEY DATA

50 USD per million tokensGPT-6 Astra output price
0.5 USD per million tokensGPT-6 Luna output price
0.1 USD per million tokensHaiku 5.5 input price
0.5 USD per million tokensHaiku 5.5 output price

PHYSIX OBSERVATION

Big tech is pulling AI back from showmanship to business: frontier models keep burning cash, while everyday tasks are handed to cheap small models. The price gap is a hundredfold, but users only pay for results. The key is who can wire up organizational data and workflows so intelligence actually saves cost, otherwise no matter how cheap it is, it is still a losing deal.

Source: TMTPost report