Claude Opus 5.5 vs GPT-6: which model is the real rival, and which one should you use?

Tuesday 22 September got my full attention: Anthropic released Claude Opus 5.5, and about ninety minutes later OpenAI answered with two new models, GPT-6 Sol and GPT-6 Luna. Most of the coverage put Opus 5.5 next to Sol and called it a price war.

Sol and Luna are OpenAI's middle and small models. The model that really competes with Opus 5.5 is GPT-6 Astra, OpenAI's top model, which came out on 3 September. On the results Anthropic published, the two trade places, with Opus 5.5 ahead on coding and everyday office work and Astra ahead on business workflows and science tasks. Keep in mind these are Anthropic's own numbers, and every lab publishes the tests it wins.

It helps to line up both families by price to see the whole picture. At the top, Claude Fable 5.1 and GPT-6 Astra cost exactly the same, $10 per million input tokens and $50 per million output tokens. Claude Mythos sits at that price too, but it is only available by invitation, so for most businesses it is not an option. In the middle, Claude Sonnet 5 and GPT-6 Sol both cost $2 and $10, and at the small end GPT-6 Luna, at $0.10 and $0.50, is ten times cheaper than Claude Haiku.

For me, the price is where this story gets interesting. Opus 5.5 costs $4 and $20, which is 60% less than Astra, and Anthropic says it performs at the level of Fable 5.1 on most work. Sol does cost half of Opus 5.5, but it plays in a lower league: on Zapier's business-workflow test it scored 33% against 40% for Opus 5.5.

So which one should you actually use in your business? I would pick the model by task and happily use whatever provider makes sense for each one. For long coding jobs and everyday office work, Opus 5.5 gives you top-level results for less money. For science-heavy work and complex business workflows, Astra is still worth testing. For high-volume simple work, like sorting emails or pulling data out of documents, I would test Luna or Claude Haiku on twenty real examples first, because at that volume the price difference adds up fast. Whatever you pick, cache the instructions you send every time, because both companies now charge ten to twenty times less for text the model has already seen.

Anthropic says smaller Claude 5.5 models are coming in the next few weeks, so the middle of the market is where the next round of this price war will be fought.