ChatGPT Sol takes on Claude Fable 5. The new battle isn’t about chatbots but about the cost of digital labour

The difference is also evident in the prices. Sol costs $5 per million tokens for incoming transactions and $30 for outgoing ones. Fable 5 charges $10 and $50 respectively. For a company that processes hundreds of millions of tokens every month, the difference becomes significant. A project that costs around $1,100 with Sol can exceed $2,000 with Claude. The price, however, tells only part of the story. If one model completes a task on the first attempt whilst another requires more human intervention, the final cost can be completely different.

Both have their weaknesses. OpenAI is paying the price for the growing complexity of its offering. With its range of models, operating modes and tools, the platform risks losing the simplicity that had been the key to its initial success. Anthropic, on the other hand, has been criticised for overly strict security filters, particularly in areas such as cyber security and scientific research, where some users have complained about a lack of transparency regarding these restrictions.

The most interesting change, however, concerns the business model. The first era of artificial intelligence was based on premium subscriptions and APIs sold to developers. Today, the focus is shifting towards services for businesses. OpenAI is complementing its licences with consultancy teams to help integrate the agents into business processes. Anthropic is strengthening its presence in the major cloud platforms and offering pay-as-you-go options for its most advanced models. Artificial intelligence is looking less and less like software to be purchased and more and more like infrastructure to be managed.

Meanwhile, new competitive pressure is mounting. Open-source and Chinese models are rapidly closing the quality gap by offering lower prices. If this trend continues, artificial intelligence risks becoming a commodity. In that case, the competitive advantage will not lie in having the best model, but in the ability to build services, applications and customer relationships.

In short, the real challenge is no longer to prove who has the most ingenious algorithm. It is to convince businesses that a digital worker generates more value than they cost. As in every industrial revolution, it is not the most sophisticated machine that ultimately wins. It is the one that makes work more cost-effective.