Opus 5 is a really good model.

Really wish this video was like 4 minutes long 😒

It can be if you play it really really fast

If i knew your subscription is so cheap I would have subbed long ago lol 😆

I think GPT 5.6 Sol is still the better choice

7 words in the tweet, 44 minutes in the video. the compression ratio of a true content creator

Do it Theo , follow me

Opus 5 is extremely lazy for me. Sol doesn't stop until it's done. Opus I have to babysit and endure it's quirky personality

"not sponsored btw"

Every single model released YTD has been virtually the same model. What we gain in intelligence we lose in models getting lost in stupid details and outright lying. Models have never been lying this much

I think it would be awesome if you added videos to your X profile showcasing T3 Code. There are many hidden features and top-tier workflows users don’t know about yet. It would be great to explore them and learn why T3 Code offers one of the best agentic coding experiences today.

Opus 5's been solid, especially for stuff that needs real reasoning depth or longer context without losing the thread. What's your use case been?

You can Run , but cannot hide. Fable is still the King.

It would be weird if Opus 5 were worse than Opus 4.8. And the fact that it outperformed Fable 5 on a few benchmarks seems more like a great PR opportunity for another round of Anthropic hype. Am I the only one tired of reading about a new 'world's best AI model' every single week? It's just a hype cycle.

Somewhere, a GPU just heard the phrase one more benchmark.

For anyone who wants the gist in 2 mins, here's a summary

Does it get flagged in cyber tasks as much as fable or is that usable?

Its probably a version of Opus maxxed and trained by Mythos :)

I’m still using GPT 5.6 Sol because with Claude I felt lot of time usage issues after buying the subscription so I’ll continue with 5.6 Sol only!

theo are you aware that they sandbagged their deep swe score?

"when I read 56's code, I question if I ever want to read code again", 💯 timestamp 38:56

Hey @theo why didn’t u address @OpenAI’s backend issue which affected millions of users today

Yeah, the positioning of fable seems a bit strange right now. What’s the next move? Should they just go ahead with fable 5.1 ? or there will be no fable at all until the Mythos come out. The whole marketing strategy is mass now for Anthropic.

classic setup for another tech breakdown

Seeing Opus 5 top the intelligence index charts is impressive.

bad hoodie, sorry

can someone link me the chinese version please?

theo has a new man chick now.

so when is it coming to t3 code?

Agreed. Opus 5 feels like the first model that actually makes multi-agent setups practical for everyday use without burning through tokens too fast. Have you noticed any big differences in how it handles long-context tasks compared to 4.8?

someone tell me, it's real the same with Fable 5?

The coding quality surprised me — it just gets stuff right more often than not.

Great deep dive as always Theo! 🙏

Anyone who thinks Opus 5 is better than Opus 4.8 clearly isn’t a developer. The evals don’t capture how terrible it is to work with Opus 5.

yeah, really good. I would still ask it to rewrite whatever Opus wrote in the last 3 days 😂