On September 22, two of the world’s biggest AI labs almost simultaneously released new models just 90 minutes apart. And both went for lower prices.
⏺ Anthropic introduced Claude Opus 5.5. The company says the model set a new record in coding and office tasks, outperforming its larger Fable model in several benchmarks. At the same time, generation got cheaper: $20 per million tokens vs. $25 for Opus 5.
💬 90 minutes later, OpenAI responded. GPT-6 Sol and Luna launched at roughly half the price of the 5.6 lineup. Sol costs $2/$10 per million tokens, down from $4/$20. OpenAI says the model makes half as many errors as its predecessor and outperforms Anthropic’s top models in several benchmarks.
💬 The race is accelerating. Opus 5 launched just two months ago on July 24. Anthropic is already promising Sonnet 5.5 and Haiku 5.5 “in the coming weeks.”
⏺ And all of this is happening amid growing discussions about the need to slow down AI development. But the market is moving in the opposite direction: models are getting more capable, releases are becoming more frequent, and prices are falling.
What’s changing❔
The AI race is gradually shifting from “Whose model is smarter?” to “Who can build a profitable business around increasingly cheaper tokens?”
For users, that’s good news. For businesses and investors, the economics are much more complicated.
📈 Follow the AI race and trade assets from the sector with TonTrader 🚀
🔗 Chat • X • TonTrader