Update — June 18, 2026: This article corrects the prediction in our earlier guide, Claude Sonnet 4.8 Is Next: Builder Preview for the June 16–18 Drop. The window has passed. Sonnet 4.8 has not shipped.
On June 6, this site published a preview article predicting that Claude Sonnet 4.8 would arrive in the June 16–18 window. The reasoning was pattern-based: Opus 4.8 launched May 28, Anthropic’s historical gap between Opus and Sonnet is roughly three weeks, and that arithmetic pointed to the week after the June 15 model-retirement deadline.
It is now June 18. There is no claude-sonnet-4-8 in the Anthropic API. There has been no Anthropic announcement, no model card, no benchmark release, no API ID. Claude Sonnet 4.6 remains the current production Sonnet, unchanged since February 17, 2026.
This article corrects that prediction, explains why the pattern broke, and gives builders an updated picture of what to plan around.
What the Current State Actually Is
The Claude 4.x Sonnet family has not moved since February:
| Model | Status | Model ID | Price |
|---|---|---|---|
| Claude Sonnet 4.6 | Available — current | claude-sonnet-4-6 |
$3/$15 per MTok |
| Claude Sonnet 4.8 | Not released | — | — |
The Opus track moved significantly since then: Opus 4.7 shipped April 16, Opus 4.8 shipped May 28. The Sonnet track skipped both. There is a two-generation gap between the current Sonnet and the current Opus — a gap that has not appeared in Anthropic’s prior release cadence.
If your production code is on claude-sonnet-4-6, you are on the correct and only available Sonnet. Nothing changed on June 15 for builders who had already migrated.
Why the Prediction Missed
The evidentiary basis was thin
The “Sonnet 4.8 is coming” narrative rested almost entirely on a single signal: a string in a leaked source-map file that accidentally shipped inside the @anthropic-ai/claude-code npm package on March 31, 2026. The file contained what appeared to be internal model references — including claude-sonnet-4-8.
That single string, combined with the Opus→Sonnet release pattern, was enough to make a reasonable probabilistic case for a June window. It was not a confirmed product signal. Anthropic never acknowledged the leak or confirmed the model ID. The article’s editorial disclaimer said exactly this.
What the pattern analysis missed: the string may have been a planning artifact, not a product commitment. Internal model strings in tooling can refer to models under evaluation, A/B test variants, or capability probes that never ship publicly. The npm leak exposed an internal snapshot, not a release schedule.
The versioning pattern broke this cycle
Anthropic has historically released matching Sonnet and Opus versions within weeks of each other. Sonnet 4.5 and Opus 4.5 shipped in the same month. Sonnet 4.6 and Opus 4.6 followed the same pattern.
This cycle broke that. Opus 4.7 shipped April 16. No Sonnet 4.7 appeared. Opus 4.8 shipped May 28. No Sonnet 4.8 appeared. Anthropic has now shipped two Opus generations without a corresponding Sonnet update — the longest Sonnet gap since the 4.x generation launched.
The most likely reading: the Mythos-class work (Fable 5, Mythos 5, and the compliance and international access negotiations around them) consumed the engineering and release capacity that would otherwise have driven the Sonnet cadence. Sonnet is not the priority when the frontier-tier is in crisis.
The Fable 5 disruption
On June 12, Anthropic suspended Claude Fable 5 and Mythos 5 to comply with a US Commerce Department export-control directive. Senior technical staff traveled to Washington for direct negotiations. That intervention — and the ongoing effort to restore international access to the Mythos tier — represents a substantial organizational focus that displaces routine product releases.
Whether this directly delayed Sonnet 4.8 is not confirmed. What is clear: Anthropic’s June has been dominated by the Fable 5 crisis, and a quiet Sonnet drop in this period would have been anomalous.
Revised Timeline Estimate
Based on the 141-day historical gap between Sonnet 4.5 (December 2025) and Sonnet 4.6 (February 17, 2026), a similar cadence from Sonnet 4.6 points to early-to-mid July 2026 as the next plausible Sonnet release window.
That estimate carries substantial uncertainty because:
- Anthropic has already broken the Opus/Sonnet synchronization pattern this cycle
- The Fable 5 negotiations have no confirmed resolution date
- It is not confirmed whether the next Sonnet will be numbered 4.7 (matching Opus 4.7) or 4.8 (matching Opus 4.8 and skipping 4.7 on the Sonnet track)
The honest answer: Sonnet 4.8 could ship in mid-July, late July, or later. It may arrive as Sonnet 4.7 instead. Builders should not build plans around a specific date.
What Builders Should Do Now
If you have already migrated to claude-sonnet-4-6: Nothing to do. You are on the current production Sonnet and the June 15 retirement deadline is behind you. When a new Sonnet ships, it will be a one-line model string change.
If you are evaluating whether to hold on claude-sonnet-4-6 or upgrade to Opus 4.8 while waiting: The answer is the same as it was in June: use Sonnet 4.6 for high-volume, cost-sensitive workloads. Use Opus 4.8 for agentic coding sessions, code review pipelines, and workloads where the alignment improvements — 4x fewer unreported defects, 17x fewer dishonest agentic summaries — justify the 1.7x price premium.
The features expected for Sonnet 4.8 — Dynamic Workflows, effort control — are real features on Opus 4.8 today. If your pipeline needs them now, test them on Opus 4.8. The API surface will be identical when Sonnet inherits them.
If you were waiting for Sonnet 4.8 to make architecture decisions: Stop waiting. Plan around claude-sonnet-4-6 as the Sonnet baseline. Treat “Sonnet 4.8 / 4.7” as an upgrade when it ships — not as a dependency.
On budget_tokens: The open question from the June 6 article — whether Sonnet 4.8 drops explicit thinking budgets in favor of effort control — remains open. Sonnet 4.6 still supports budget_tokens. If your pipeline depends on this, Sonnet 4.6 is the stable choice until Anthropic confirms the behavior on the next Sonnet.
The Corrected Decision Matrix
Until the next Sonnet ships:
| Workload | Recommended model |
|---|---|
| High-volume pipelines: classification, extraction, summarization | claude-sonnet-4-6 |
| Cost-sensitive production at scale | claude-sonnet-4-6 |
| Latency-sensitive tasks | claude-sonnet-4-6 |
| Long-horizon agentic coding, codebase-scale review | claude-opus-4-8 |
| Workloads needing Dynamic Workflows now | claude-opus-4-8 |
Pipelines with explicit budget_tokens dependency |
claude-sonnet-4-6 |
This matrix will update when Anthropic announces the next Sonnet. Sign up for Anthropic’s release notes at platform.claude.com/docs/en/release-notes/overview to catch the announcement on the day it ships.
A Note on Prediction-Based Coverage
Predicting model release windows from pattern analysis and leaked strings is genuinely useful to builders who need to plan. It is also imprecise by design — the signal is indirect and the error bars are real. When a prediction misses, saying so clearly is more useful than leaving the original article uncorrected.
The June 6 article included a disclaimer that Sonnet 4.8 had not been announced. The disclaimer was accurate. The prediction in the title and summary was wrong. This article is the correction.
When Anthropic announces the next Sonnet — whether numbered 4.7, 4.8, or otherwise — this site will cover it the same day.