OpenAI says its latest model has overtaken Anthropic.

OpenAI says its latest model has overtaken Anthropic. But what does that actually mean for the people using it?

GPT-5.6 Sol has moved ahead of Claude Fable 5 on OpenAI’s reported coding-agent benchmark. According to OpenAI, it completed the work in less than half the time, used fewer than half the output tokens and cost around one-third less.

That does not mean it will be better than Claude at everything. Benchmarks are useful indicators, not guarantees. But the improvements point towards something commercially important: more capable AI that needs less supervision to complete complex work.

So, what can we do with it now?

We can give the model a broader outcome rather than one small instruction at a time. It can research a subject, inspect documents, work across files, use software tools, write or improve code, test its own work and carry a task through several stages.

For a business, that could mean:

• researching a market and turning the findings into a useful report
• analysing customer information and identifying patterns
• creating and improving website content
• building internal tools and automating repetitive processes
• reviewing an ecommerce journey and suggesting practical improvements
• helping a team move from an idea to a tested first version much faster

The biggest change is not simply that the model gives better answers. It is becoming better at doing the work around the answer. Important facts, calculations and decisions.