Anthropic Releases New AI Model, Claude Opus 4.8 Anthropic’s latest AI upgrade, Claude Opus 4.8, aims to be faster, cheaper, and—crucially—more candid about its own mistakes, even as the company withholds an even more powerful system, Mythos, over safety concerns.

Early May: Mythos in the wings

Anthropic first positioned Mythos as its most advanced model, but limited access to “just a handful of partners” amid cybersecurity worries. The company signaled that broader Mythos‑class access would follow once additional safeguards were in place.

May 28: Anthropic unveils Opus 4.8

On May 28, Anthropic announced “Introducing Claude Opus 4.8,” describing it as an upgrade to Opus 4.7 with better benchmark performance and collaboration, available “for the same price.” The company highlighted new “effort control” in claude.ai, a cheaper “fast mode” that runs at 2.5x speed, and “dynamic workflows” in Claude Code to tackle very large-scale coding tasks with swarms of subagents.

Tech outlets quickly echoed the focus on honesty. The Verge reported that Anthropic trains models “to avoid making claims that they can’t support” and that early testers found Opus 4.8 “more likely to flag uncertainties about its work and less likely to make unsupported claims,” roughly “4x less likely than its predecessor” to let its own code flaws pass unnoticed. TechCrunch similarly stressed the short, 41‑day jump from Opus 4.7 and said early testers found the model “more likely to flag uncertainties about its work,” especially with bad or uncertain data.

Axios framed the release as a value play—better coding and knowledge‑work skills “for the same price” and a fast mode that’s three times cheaper than before—while noting Opus 4.8 “still lags the performance of Mythos,” which is promised “in the coming weeks.”

User and industry reactions

Anthropic’s own system card and testimonials claim improved judgment, tool use, and reliability across coding, legal, and research applications. The Next Web called Opus 4.8 “four times more honest,” citing alignment tests that show lower rates of deceptive or misuse‑enabling behavior compared with Opus 4.7.

Online, reactions mixed seriousness with humor. Elon Musk responded with a single “😂” to a viral post joking, “Me using Claude Opus 4.8 to rename a file,” hinting at skepticism—or amusement—about the fanfare around incremental upgrades.

Continue reading https://foxvector.com/stories/019e7f06-6f4d-13af-71fd-0f75e7f38e45