# Opus 4.8: the benchmarks crept up, but the fast mode is the actual upgrade

> Opus 4.8 landed on 28 May 2026 with modest benchmark gains but a genuinely useful fast mode.

*Five points on a coding benchmark. A 2.5× speed-up. Guess which one you'll feel.*

By The InsidersFeed Desk · InsidersFeed
Canonical: https://insidersfeed.com/news/opus-4-8-fast-mode-is-the-upgrade

> **Key:** **The take:** the headline benchmark bump is a few points — nice, not seismic. The stuff you'll actually feel is the **2.5× fast mode** and the long context. And the tease at the bottom of the post mattered more than the whole model.

Anthropic dropped **Opus 4.8** on 28 May 2026, about six weeks after 4.7. Same six-week heartbeat, same polished blog post. Let's separate the marketing from the upgrade.

## The benchmark, in perspective

Anthropic's own number: agentic coding **64.3%→69.2%**. That's a five-point move on Anthropic's chosen metric — genuinely useful, not a revolution, and (as always) a figure the vendor picked and you should treat as marketing-adjacent. More interesting is the reliability claim: **~4× less likely** to let flaws in its own code pass unremarked. If that holds up in real agent runs, it's worth more than any leaderboard.

> **Note:** **Credit where due:** a 1M-token context plus a 2.5×-faster, cheaper 'fast mode' is a real, shippable improvement for anyone working over big codebases — that's not hype, that's a better tool. We're knocking the framing, not the product.

## The part they buried

At the bottom, the post teased **'Mythos-class'** models 'in the coming weeks.' Read that as: *4.8 is the appetiser.* Days later, Fable 5 — the first public Mythos model — showed up, only to get yanked offline by a US directive. So the honest summary of Opus 4.8 is that it was a solid, incremental release whose most important line was a pre-announcement of the model that actually made the news.

## FAQ

### Is Opus 4.8 a big jump over 4.7?
Incrementally. Anthropic's headline coding figure moved 64.3%→69.2% — useful, not dramatic. The practical wins are the ~2.5× fast mode and 1M-token context, plus a claimed 4× drop in missed code flaws.

### Should I switch to the fast mode?
If you're iterating over large codebases and want cheaper, quicker turns, it's the upgrade worth trying. The benchmark delta matters less day-to-day than the speed and context do.