Anthropic isn’t but able to launch its supposedly super-powered Claude Mythos AI mannequin to most people. Nevertheless, the AI firm simply launched an improve to its flagship product, Claude Opus, which is now at model 4.8.
“Based mostly on Opus 4.7 with enhancements throughout the benchmarks to make it a more practical collaborator,” says Anthropic. promised in the press release Thursday. In truth, the benchmark numbers beneath present a really small enchancment total.
One of many huge enhancements is reportedly within the space of hallucinations. Claude Opus 4.8 would not deceive its customers that a lot. “Early testers report that Opus 4.8 is extra prone to flag uncertainties about its operation and fewer prone to make unsubstantiated claims,” Anthropic stated, stressing the mannequin’s “integrity.”
Claude works 4.8 has “higher judgment”
“Claude Opus 4.8’s judgment is considerably higher,” Shopify engineer Tom Pritchard advised Anthropic. A coding model of this mannequin is to “ask the correct questions, discover your errors, and push again in case your plan is not the correct one.”
That promise could also be music to the ears of vibe coders all over the world, given the rising variety of horror tales about AI brokers deleting total company databases.
mashable gentle velocity
To fulfill energy customers, Anthropic is closely discounting a “quick mode” that permits Claude to run 2.5 occasions sooner than regular. Quick mode is “3 times cheaper than earlier fashions,” the corporate stated.
Reddit users didn’t buy ithowever. Many have been involved that they might lose entry to the extra well-liked mannequin, the Claude Opus 4.6. “No one trusts benchmark charts,” one redditor summed it up, noting that Opus 4.7 additionally appeared to have fairly good numbers on the time of launch.
Whether or not or not you possibly can belief the benchmarks, thoughts you, Mashable hasn’t independently verified these numbers, however this is what Anthropic claims:
Credit score: Anthropic
The way to attempt Claude Opus 4.8
Claude Opus 4.8 is offered now from Anthropic’s web site. Claude.AIout there through the Claude API in addition to through Anthropic companions comparable to Microsoft Foundry.
The brand new mannequin is exactly the same price That’s, fashions courting again to Claude Opus 4.5. All of this prices $5 per million enter tokens and $25 per million output tokens.
Nevertheless, provided that Anthropic is promising a Claude mythos within the coming weeks, it is perhaps price ready a bit to see if the mannequin will be much more “sincere” about its hallucinations.
Subjects
synthetic intelligence anthropology
