Calvin Wankhede / Android Authority
TL;DR
- Anthropic has released its latest flagship model: Opus 4.8.
- It performs better than Opus 4.7 across all benchmarks, and has a cheaper ‘fast mode’ than previous models.
- Claude is also getting ‘Dynamic workflows’ in research preview, and new effort controls.
If you’re a Claude user, there are several new features and improvements headed your way. A new flagship model, effort controls, and a lot more are now available.
Anthropic today announced the latest version of its flagship model, Claude Opus 4.8, alongside several new features. The new model builds upon the strengths of Opus 4.7 and delivers better benchmark scores in almost every aspect.
Don’t want to miss the best from Android Authority?
The company said that Opus 4.8 “reaches new highs on our measures of prosocial traits.” It also claimed that the new model is more inclined to support user autonomy and act in the user’s best interest. Further, it has a significantly lower chance of being deceptive or cooperating with misuse.
It’s also more honest, per Anthropic, and is more likely to highlight when it’s uncertain about something rather than jumping to conclusions or making unverified claims about its work. The company’s evaluations also show that Opus 4.8 is four times less likely to not report flaws in code it writes.
For users who are worried about the rising costs of AI use, Opus 4.8 offers some good news. Using Opus 4.8’s fast mode for faster responses is now three times cheaper than before. That means you can use the model for simpler queries as well, without worrying about it draining your tokens too quickly.
Apart from the new model, Anthropic has also introduced new features for Claude.
Users can now control how much effort Claude puts into a task. A new ‘Effort’ option in the model selector lets users choose between Low, Medium, High, and Max.
With lower effort settings, Claude will offer faster results, while higher settings will make the AI think longer and offer more detailed replies. It’s also worth noting that higher effort settings will burn through tokens much faster, and the company recommends using the ‘Max’ effort mode for the hardest tasks only.
A new ‘Dynamic workflows’ feature is also available in research preview and can allow users to work on bigger problems in Claude Code. Just give it a task and it can plan everything out and work on it by running hundreds of subagents simultaneously. Claude will even verify the output before showing it to the user.
Claude Opus 4.8 is available everywhere starting today, and the company is also planning on releasing Mythos-class models in the coming weeks.
Thank you for being part of our community. Read our Comment Policy before posting.
