Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- Opus 5 targets coding, knowledge work, and agent workflows.
- Anthropic says it uses fewer tokens for similar quality.
- Developers get faster output and smarter tool controls.
To the surprise of no one, Anthropic has released Claude Opus 5. After releasing the controversial Mythos 5 and Fable 5 last month and Sonnet 5 on July 1, it was inevitable that Opus would be bumped from 4.8 to 5. That’s happening today.
I’ve been impressed with Opus 4.8. I’ve been using it for a number of large coding projects, and it’s handled them quite well. But based on limited user reports, it’s likely that Claude Code will get another big boost.
Also: OpenAI’s attack agent did exactly what it was told – just more relentlessly than expected
I’m not the only one impressed with Opus, based on the accolades included in Anthropic’s announcement.
Cristian Rivera, software engineer at Stripe, says, “Opus 5 is a clear generational step up from Opus 4.8. Over one weekend, I gave it a chief-of-staff role over my dev environments: it built its own monitor, drove each box, and pulled me in only for the judgment calls.”
Harvey creates AI legal tools. According to Niko Grupen, Harvey’s head of applied research, “Claude Opus 5 is a clear step up in performance on legal agent work compared to prior Opus models, and we saw the biggest gains in practice areas like corporate governance and arbitration.”
He adds: “We were also impressed with Opus 5’s ability to maintain quality at lower reasoning levels, achieving similar performance while generating 26% fewer tokens on average compared to Opus 4.8 at max reasoning.”
Token use isn’t just a measure of cost. It’s also a measure of how much stress the AI is handling while it’s working on complex issues. There have been a number of times, even with Opus 4.8’s million-token context window, that my Claude Code has requested a recess and to take time to nap.
Using fewer tokens for the same work means it might go farther in a given session. That would be helpful.
Near Fable 5 performance
Anthropic says that “Opus 5 comes close to the capabilities of Claude Fable 5 in many domains at half the price.” The company reports that it has the strongest knowledge work performance of any model it’s released, and delivers substantial gains on complex coding. These two would go directly to improve work in both Claude Cowork and Claude Code.
Also: Claude Fable 5 is back, but I’m sticking with Opus 4.8 for daily work: 5 reasons why
AI development vendor Cursor supports the assessment that Opus 5 is nipping at the heels of Fable 5 in terms of capabilities. Cursor co-founder Sualeh Asif reports, “Claude Opus 5 delivers near Fable 5 intelligence at Opus speed and cost. On CursorBench, it’s just under Fable 5 and has many of the same behaviors. We are excited to see how developers use it in Cursor.”
Anthropic says it expects Opus 5 to be the daily driver, especially in the enterprise. I’ve been using the heck out of Opus 4.8 on my $100/month Claude Max plan and haven’t experienced throttling. I spent the last week putting it through a lot of big agentic runs and am still at only 31% of my usage limit for the week, which resets later today.
Presumably, I’ll move to Opus 5 as my go-to model and will probably have even more runway to accomplish complex tasks.
Also: I used Claude Code to vibe code a Mac app in 8 hours, but it was more work than magic
While Opus 5 is better for knowledge work (that’s kind of table stakes for a major version upgrade), it also offers greater autonomy with less back-and-forth, better error recovery, and more frequent checks of its work.
What I find particularly interesting is that Anthropic is again mentioning how Opus is helping with biology and other scientific research. The company says that Opus 5 “overperforms on biology tasks compared to Opus 4.8.” Unfortunately, the company hasn’t yet defined “overperformed” in this context, but we can fairly assume it means better, faster, and cheaper than Opus 4.8.
Anthropic says the pricing is the same as for Opus 4.8. For token API users, it’s $5 per million input tokens and $25 per million output tokens. For Claude Max users like me, it will be the default model offered and will have “standard limits on all paid plans.”
Also: Claude Fable 5 is back, but I’m sticking with Opus 4.8 for daily work: 5 reasons why
Anthropic still recommends using Fable 5 for “your most ambitious work, the days-long autonomous projects nothing could take on before.” I’m getting ready to add the in-app purchase wiring for my iPhone and Mac apps, and I’m thinking I might assign that big (and scary) project to Fable to see what it can do for me.
Other announcements
The company is also making a series of related announcements today. Here’s a quick lightning round.
- Fast mode for Opus 5 on the Claude platform: This is a research preview that enables full Opus 5 capability with 150% faster token generation at double the standard price. For subscription Max users of Claude Code, this feature can be bought with extra usage credits.
- Automatic fallbacks: Anthropic says, “When a safety classifier declines a request, the API seamlessly falls back to another model within the same request, so your users get a helpful response instead of an error.”
- Mid-conversation tool changes: Developers can swap out which external tools Claude is using in the middle of the conversation, without nuking the prompt cache. This allows developers to turn off tools once they’re no longer needed, helping both with cost control and security.
Opus 5 is available now to all user tiers that previously had access to Opus 4.8.
Would you use Claude Opus 5 as your daily AI coding model? Let us know in the comments below.
You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.
