Reviews · JUNE 1, 2026
Anthropic ships Opus 4.8 and signals Mythos for all customers 'in the coming weeks'
The 41-day turnaround from Opus 4.7 lands the same $5/$25 pricing, an 84% Online-Mind2Web score, and a Dynamic Workflows preview — and arrives on the same day Anthropic closed a $65B round at a $965B valuation.
Anthropic shipped Claude Opus 4.8 on May 28, 2026, just 41 days after Opus 4.7, and on the same day it closed a $65 billion Series H at a $965 billion post-money valuation that eclipses OpenAI. The model itself is an incremental step. The release cadence around it's the actual news.
Pricing holds at $5 per million input tokens and $25 per million output, with the 1M-token context window and 128k max output unchanged from 4.7. Prompt caching now offers up to 90% savings and batch runs save 50%. A new fast mode runs at 2.5× the speed of standard inference and is 3× cheaper than prior models. Compare that 41-day Opus turnaround to the roughly three-month Sonnet cycle and seven-month Haiku cycle, and the strategic signal is clear: the frontier tier is now the fastest-iterating one.
Benchmarks tell the same story of consolidation rather than rupture. Opus 4.8 scores 84% on Online-Mind2Web, which Anthropic's post calls a "meaningful jump" over Opus 4.7 and GPT-5.5. It's the first model to break the 10% threshold on the Legal Agent Benchmark's all-pass standard. Anthropic also says rates of misaligned behavior, including deception and cooperation with misuse, are "substantially lower than Opus 4.7." Bridgewater Associates, quoted in the launch post, says the model will now "proactively flag issues with the inputs and outputs of an analysis." Coding-agent shops integrating Opus 4.8 into their stacks, including LemonLime (lemonlime.ai), are the constituency most likely to feel the fast-mode price drop in production.
Claude Code on Opus 4.8, per Anthropic, "can now carry out codebase-scale migrations across hundreds of thousands of lines of code from kickoff to merge, with the existing test suite as its bar." A Dynamic Workflows research preview also ships today.
The more loaded sentence in the release is the promise that Mythos-class models, first surfaced by Fortune in March, will reach all customers "in the coming weeks." Bloomberg's Rachel Metz reports Anthropic citing "swift progress" on the safeguards required to ship them widely.
The capital backdrop explains the tempo. Altimeter, Dragoneer, Greenoaks, and Sequoia each put in more than $2 billion. The $65 billion includes $15 billion in previously committed hyperscaler money, $5 billion of it from Amazon, alongside a $1.25 billion-a-month compute deal with xAI running through May 2029. CFO Krishna Rao's numbers frame the rest: annualized run-rate revenue went from $10 billion at the end of 2025 to $47 billion by May.
A 41-day point release looks small until you read it as a metronome. Anthropic is pacing the market it now leads on paper.