A source map accidentally shipped with the @anthropic-ai/claude-code npm package v2.1.88 on March 31 contained three strings in a security-filter list: sonnet-4-8, opus-4-7, and mythos. With Opus 4.7 having shipped on schedule, the other two strings now carry real weight.
No model card, announcement, or API ID exists yet for Sonnet 4.8 — but the leak’s track record makes a mid-June release look close to certain.
Why the leak is credible
The same source map listed opus-4-7, and Opus 4.7 subsequently shipped on April 16 — exactly as the string implied. A leak that correctly called one unreleased model gains credibility for the others alongside it.
| Leaked string | Status |
|---|---|
| opus-4-7 | ✅ Shipped April 16 |
| sonnet-4-8 | ⏳ Expected mid-June |
| mythos | ❓ Unidentified |
The skipped-version puzzle
Anthropic has never skipped a minor version. Jumping from Sonnet 4.6 to Sonnet 4.8 — with no Sonnet 4.7 — would be unprecedented. The likeliest explanation is version alignment: matching Sonnet’s number to the Opus 4.8 line so the family shares a generation.
What Sonnet 4.8 likely improves
Expect trickle-down from Claude Opus 4.8: stronger coding and agentic performance, and improved vision. Sonnet’s role is the high-volume workhorse, so the gains that matter most are the ones that hold up across millions of cheaper, faster calls.
Why developers should care
Sonnet 4.6 is the workhorse behind Perplexity’s orchestrator, Claude Code’s lighter tasks, and millions of daily API calls. A 4.8 jump changes the cost-performance math for production: if it narrows the gap to Opus on coding while keeping Sonnet pricing, many workloads currently on Opus could move down a tier.
For teams choosing a production model now, the practical move is to hold benchmark plans ready — if Sonnet 4.8 ships mid-June as the leak implies, re-test it against your current Sonnet 4.6 and Opus 4.8 workloads before committing.
