Claude Opus 5 Rumors Are Here - Anthropic Is in Trouble

Rumors about Claude Opus 5 just started to surface, and I am here to tell you that this is not necessarily good news. In this breakdown I map out the exact situation Anthropic is walking into: a departing Fable 5, a GPT 5.6 Sol that already sets a brutal benchmark, and a zone of outcomes I call the valley of death, where Opus 5 could land and still lose.

1. The Claude Opus 5 rumors (0:00)

Claude Opus 5 is now available on Vertex, and there are quite a few places hinting at a release sooner rather than later. On its own, that would be exciting. But you have to read it against the current situation between Anthropic, OpenAI, and SpaceX AI, and that situation is uncomfortable.

2. What happens when Fable 5 goes away (0:24)

The core problem: Fable 5 is theoretically going away on July 19th, just a couple of days from when I recorded this. Take Fable 5 out of the market and, systematically, GPT 5.6 Sol becomes the best model out there. That is the world Opus 5 will be born into. It does not get to compete against a friendly baseline; it has to outperform everything GPT 5.6 Sol already does, and that is very, very difficult.

3. The Opus 5 valley of death (0:50)

So I defined something I call the valley of death for Opus 5. Picture the line between Opus 4.8 and GPT 5.6 Sol. If Opus 5 performs anywhere inside that line, it is a useless model. Why? Because you are simply better off going to OpenAI. They give you more usage, and they are not telling you they will take away your model at any point in time, which quite honestly sucks. We did not even get a usage reset for the people who aligned their usage to match July 12th, the supposed end date for Fable.

Anywhere above GPT 5.6 Sol is decent, but not automatically good. A lot of people will still be angry that the best model is not the one included in the subscriptions they are paying for. And even in the success case, Anthropic still has to find a way to make Fable 5 available at least to people on the $100 plan. Anywhere below 5.6 is death valley. It is a crucial moment for Anthropic.

4. Why beating GPT 5.6 Sol may not be enough (1:20)

Beating Opus 4.8 is nearly not enough. They have to beat GPT 5.6 Sol, and they have to beat it by a big, big, big margin. If not, things are going to get very annoying for them: a narrow win does not change anyone's subscription decision when the competitor offers more usage with fewer threats of removal.

5. The performance people expect from Opus 5 (2:07)

There is a second trap. A lot of people expect Fable 5 levels of performance from Opus 5, and that is not at all guaranteed, because Fable 5 is another type of model. It is not the same type of brain as Opus, so that bar is going to be very hard to reach. People also expect to use the new model forever, with no sunset hanging over it. If Anthropic does not deliver on those two expectations, performance and permanence, it is going to be quite rough. The only thing left for us to do right now is wait and see what Opus 5 is truly capable of.

6. OpenAI's next move (2:46)

Meanwhile, OpenAI is still working on a new model. As usual, that is their job. We might see another model coming from OpenAI within the same timeframe as Opus 5, which would move the goalposts yet again just as Anthropic tries to clear them.

7. GLM and xAI tighten the race (3:12)

And it is not just OpenAI. There is GLM, the open source AI model from the Chinese company Zhipu, and SpaceX AI is working on newer models that are trailing just right behind the Opus family. The pack is close, and it is closing.

8. Why I am not excited yet (3:27)

I wish I could be excited that Opus 5 is coming soon. I am not, because I know it will not be as good as Fable 5 by default. That is not pessimism for its own sake; it follows directly from the fact that these are different kinds of models with different strengths.

9. A future with no single best model (3:47)

What this all points to is a very interesting era where there will not be one best model. There will be several different ones from different providers, each better at different things. If you want to follow the story as it develops, go look around Twitter: there is quite a bit of information on Opus 5 right now, and I expect even more over the coming two to three weeks. I really hope Anthropic figures it out.

Where to go next

• Open source is adding to the pressure from below. See Kimi K3 Just Put Claude and OpenAI on Notice.
• How this pressure already changed Anthropic's plans once: Fable 5 Gets 7 More Days: Here's Why.