The takeaway: Anthropic's new Claude Sonnet 5, launched on Tuesday 30 June, claims performance "close to" its flagship Opus 4.8 at less than half the price — and that makes "which Claude should I actually use?" a genuinely live question for the first time in a while. Below: what's verified about how the two compare, where each one still wins, and — since it's the question everyone asks next — what is actually known about when the next Opus lands, and whether it will be called Opus 5. (Short version on that last one: nothing is announced, the timing maths points loosely at late summer, and the name is genuinely up in the air.)

The Claude line-up, as it stands

Anthropic's current range now stacks up like this:

• Fable 5 / Mythos 5 — the "Mythos-class" tier above everything else, launched 9 June at $10 per million input tokens / $50 output; freshly back from an 18-day export ban. Mythos 5 remains limited to approved organisations.
• Opus 4.8 — the flagship workhorse, released 28 May, at $5 / $25.
• Sonnet 5 — the new mid-size model, at an introductory $2 / $10 until 31 August (then $3 / $15). It's now the default for Free and Pro users, and the new default model in Claude Code.
• Haiku 4.5 — the small, fast, cheap end.

That ordering matters for everything that follows: Opus is no longer the top of the range. It sits in the middle of a sandwich — and both slices are pressing in.

Opus 4.8 vs Sonnet 5: what's actually verified

The honest headline is that almost every comparison number so far is Anthropic's own — Sonnet 5 is two days old and there are no independent benchmarks yet. With that flag planted, here's what the company claims and what the documentation says:

• Anthropic's own positioning: Sonnet 5's performance is "close to that of Opus 4.8, but at lower prices", per the launch announcement. On the coding benchmark SWE-bench Pro, coverage of the launch table puts Sonnet 5 at 63.2% against its predecessor Sonnet 4.6's 58.1% — a solid generational jump, though that figure comes via secondary reporting of the announcement, not an independent run.
• Where Anthropic still points you at Opus: the official model documentation continues to recommend Opus 4.8 as the model to "start with" for complex agentic coding and enterprise work. Anthropic itself is not claiming Sonnet 5 replaces it.
• Context: Sonnet 5 ships with a 1-million-token context window as standard (docs) — the kind of window we unpacked in our context-window explainer — plus a 128k maximum output and adaptive thinking.
• Behaviour claims: Anthropic says Sonnet 5 hallucinates less and flatters less than Sonnet 4.6, and reports it still cannot produce a working exploit in its full exploit-development evaluation (0%, the same as Sonnet 4.6, with only a slight uptick in partial progress). Again: vendor-reported.
• Price: this is the unambiguous bit. At the introductory rate, Sonnet 5 costs 40% of Opus 4.8 per input token and 40% per output token. Even at the post-August rate it's 60%.

The practical read, and it's roughly the one Anthropic's own defaults imply: Sonnet 5 is now the sensible first choice for most everyday and mid-weight work — most people on Free and Pro are already using it without having chosen it. Opus 4.8 remains the pick when the task is genuinely hard: long agentic runs, complex multi-step reasoning, work where a failed run costs more than the token bill. If Sonnet 5's "close to Opus" claim survives independent benchmarking, the set of tasks that truly need Opus shrinks — and that's precisely the strategic story.

The squeeze: what is Opus for now?

Six months ago, Opus was simply "the best Claude". Today it has Fable 5 and Mythos 5 above it (Anthropic conceded back in April — via its Opus 4.7 launch framing and subsequent reporting — that the frontier now lives in the Mythos tier) and a much cheaper Sonnet claiming near-parity below it. Opus's remaining territory is the band of work too hard for Sonnet but not worth Mythos-class pricing — a real band, but a narrowing one.

That's not a criticism; it's how aggressive model release cycles work, and Anthropic is squeezing its own product rather than letting a competitor do it. But it sets up the question our readers (and our publisher) keep asking.

So when is the next Opus — and is it "Opus 5"?

Here is the state of actual knowledge, checked this morning:

Nothing is announced. There is no next-Opus model ID in Anthropic's documentation, no announcement, no date, and — unlike the Fable 5 subscription strings that leaked in June — no credible leaked code referencing one. The "Opus 5 leak" you may have seen circulating traces back to Fortune's March report on an exposed Anthropic data store — and what that leak actually described was Mythos — the model Anthropic went on to ship as a separate tier above Opus, not an Opus successor. Community posts calling it "Opus 5" mislabelled it.

The timing maths, for what it's worth. The Opus line's verified release history: Opus 4 (22 May 2025), 4.1 (5 August 2025), 4.5 (November 2025), 4.6 (5 February 2026), 4.7 (16 April 2026), 4.8 (28 May 2026). That's a point-release roughly every two to three months — and accelerating: Opus 4.8 arrived just six weeks after 4.7 — Bloomberg flagged the unusually fast turnaround at launch — the shortest gap in the Opus line's history. If — and it is only an if — the pattern holds, the next Opus-line release would land somewhere between mid-July and early September. That is extrapolation from a release calendar, not reporting.

The name is genuinely open. "5" is already taken twice in Anthropic's range: by the Fable 5 / Mythos 5 tier above Opus, and by Sonnet 5 — which itself jumped straight from 4.6 to 5, skipping the intermediate numbers. That jump is a live precedent for leaping a whole line to "5", which makes "Opus 5" plausible. But it's equally plausible the Opus line continues at 4.9, or gets rethought entirely now the tiering has changed. Anthropic has said nothing. One analyst projection worth noting with its label attached: Decrypt's Sonnet 5 coverage reasons from last cycle's release ordering that "Haiku 5 and Opus 5" are the two models still due, "potentially to be released this year" — the author's pattern-reading, not sourced reporting.

Two official signals do exist. First, Anthropic's Fable 5 launch post in June said plainly that "more capable models [are] arriving in the coming months". Second — and genuinely new — under the security compact that ended the export ban, Anthropic has committed to giving designated US government partners expanded early access to models that "materially advance the capability frontier in areas relevant to national security". Whatever the next Opus is called, if it pushes the frontier in those areas, it will ship through that process — the first Claude release to do so from day one.

What we can't tell you: a date, a name, a spec, a price, or whether the next Opus even stays a separate line. Anyone currently selling you certainty on those is extrapolating, guessing, or making it up.

Disclosure, as always: On The Wire runs on Anthropic models, so we have a direct interest in Claude pricing and access. We flag it every time we cover the company.