explainx / blog
Claude Sonnet 5 launched June 30, 2026. Now the default on Free and Pro plans. API model ID: claude-sonnet-5. Intro pricing $2/MTok input through Aug 31, 2026. Full specs, benchmarks, and what changed from Sonnet 4.6.
Jul 1, 2026
Anthropic''s official "Redeploying Fable 5" post — export controls lifted June 30, global restore July 1, stricter classifiers, 50% of weekly limits through July 7, credits after. HN breakdown: coding confusion, Opus fallback billing, invisible competitive-use limits, and Commerce reevaluation risk.
Jul 1, 2026
App strings showed Fable credits gated on verification before restore. Fable 5 is live July 1 — how credits and ID checks may still work.
Jul 1, 2026
Fable 5 is live July 1. June 13 limits reset during ban; credits may still apply. Sonnet 5 raised limits. GPT-5.6 broad access around the corner.
Update — June 30, 2026: Claude Sonnet 5 is live. Anthropic launched Claude Sonnet 5 on June 30, 2026. It's now the default model on Free and Pro plans. API model ID: claude-sonnet-5. Intro pricing: $2/MTok input, $10/MTok output through Aug 31, 2026.
| Benchmark | Sonnet 5 | Sonnet 4.6 | Opus 4.8 |
|---|---|---|---|
| SWE-bench Pro (agentic coding) | 63.2% | 58.1% | 69.2% |
| Terminal-Bench 2.1 (agentic coding) | 80.4% | 67.0% | 82.7% |
| Humanity's Last Exam (no tools) | 43.2% | 34.6% | 49.8% |
| Humanity's Last Exam (with tools) | 57.4% | 46.8% | 57.9% |
| OSWorld-Verified (computer use) | 81.2% | 78.5% | 83.4% |
| GDPval-AA v2 (knowledge work) | 1618 | 1395 | 1615 |
See the charts above for Sonnet 5's full benchmark profile across agentic coding, computer use, and knowledge work tasks. Sonnet 5 notably surpasses Opus 4.8 on knowledge work (GDPval-AA v2: 1618 vs 1615) while matching it closely on computer use.
Claude Sonnet 5 does not officially exist yet. On June 30, 2026, a wave of X posts and AI-news accounts reported that Anthropic's next Sonnet model is imminent — some naming today specifically. The evidence is real enough to track; the conclusion is not confirmed.
This article maps what is verified, what is rumor, and why the same internal slug already fooled the community once in February.
| Question | Answer |
|---|---|
| Is Sonnet 5 confirmed? | No — no Anthropic announcement, docs entry, or API listing |
| Why the buzz today? | Partner-platform slug sightings + X posts from @M1Astra, @mark_k, @AiBattle_ |
| Current Sonnet baseline | Claude Sonnet 4.6 (claude-sonnet-4-6) |
| Fable 5 status | Suspended since June 12 export-control directive |
| What to build against now | Sonnet 4.6 for mid-tier; Opus 4.8 for frontier fallback |
| Release signal to watch | anthropic.com/news + /v1/models + pricing page — all three |
The current cycle started with renewed sightings of a claude-sonnet-5 model identifier on Anthropic partner platforms — the same pattern that surfaced on Google Vertex AI logs in February 2026. On X today, several accounts amplified the story:
None of these accounts are Anthropic. The posts accumulated significant engagement (40–50+ reposts each in the first hours), which is how rumor cycles accelerate in the AI community — not how Anthropic typically announces models.
According to WaveSpeed's API watch (updated June 23, 2026), Anthropic's official models overview still lists Opus 4.6/4.7/4.8, Sonnet 4.6, Haiku 4.5, and the suspended Fable 5 / Mythos 5 entries. No Sonnet 5.
The most important historical detail: the exact same claude-sonnet-5 slug already appeared — and shipped as Sonnet 4.6.
| Date | Event |
|---|---|
| Early Feb 2026 | claude-sonnet-5@20260203 surfaces in Vertex AI logs with codename "Fennec" |
| Feb 17, 2026 | Anthropic releases Claude Sonnet 4.6, not Sonnet 5 |
| June 9, 2026 | Fable 5 and Mythos 5 launch |
| June 12, 2026 | US export-control directive; Fable 5 and Mythos 5 suspended globally |
| June 21, 2026 | claude-sonnet-5 slug reappears on partner platform |
| June 30, 2026 | X rumor cycle peaks; "today" predictions proliferate |
Internal model identifiers are staging labels. Partner platforms expose them in error logs and integration tests before public launch. That exposure is a weak signal — useful for watchers, useless as a release calendar.
As Pasquale Pillitteri's analysis notes, rumored SWE-bench scores (82–92%) and coordinated GPT-5.6 timing are third-party estimates, not Anthropic communications. The "Fennec" codename is recycled from the February leak that became 4.6.
Nine days after Fable 5 went dark, a new Sonnet slug is emotionally loaded context — not proof.
Anthropic's most capable public model was pulled for every user worldwide on June 12. Developers who had migrated workflows to Fable 5 were redirected to Opus 4.8. The community is actively looking for Anthropic's next move — whether that is a Fable 5 restoration, a new Sonnet tier, or both.
One thread on X (@kimmonismus) raised a sharper policy concern: if Sonnet 5 launches alongside a Fable 5 re-release that requires identity verification (presumably US-only or similarly restricted), users who cannot access Fable 5 would effectively get a mid-tier model as their best Anthropic option. That is speculation — but it follows logically from the export-control geography established in June.
Anthropic has not commented on any linked Sonnet 5 / Fable 5 release strategy.
A Claude model is available when all three are true:
/v1/models returns the model with pricing listedUntil then, treat Sonnet 5 as unconfirmed.
If Anthropic ships a new Sonnet this week, expect it to slot between Haiku 4.5 and Opus 4.8 — the same position Sonnet 4.6 holds today. Based on Anthropic's recent release pattern:
| Dimension | Likely direction |
|---|---|
| Agentic coding | Incremental SWE-bench uplift over Sonnet 4.6 (~79% at 4.6 launch) |
| Pricing | Same tier as Sonnet 4.6 unless Anthropic reprices the line |
| API ID | Possibly claude-sonnet-5-* or another 4.x suffix (see February precedent) |
| Effort parameter | Supported — Sonnet 4.6 already defaults to High effort; see Effort parameter guide |
| Fable 5 gap | Would not replace Fable-class reasoning; Opus 4.8 remains the official fallback |
For context on Anthropic's current flagship tier, see Claude Opus 4.8 launch coverage.
| Model | Status | Best for |
|---|---|---|
| Fable 5 | Suspended globally | Was frontier reasoning; see ban timeline |
| Mythos 5 | Suspended | Enterprise / Glasswing only |
| Opus 4.8 | Available | Frontier fallback; agentic benchmarks |
| Sonnet 4.6 | Available | Production default for coding + agents |
| Haiku 4.5 | Available | Fast, low-cost tasks |
| Sonnet 5 | Unconfirmed | Rumored only |
Sonnet 4.6 shipped with Adaptive Thinking and High default Effort — any Sonnet successor would likely inherit both. If you are tuning agent harnesses, see loop engineering for coding agents before swapping model IDs.
When Anthropic ships a model, verification is mechanical. Run this checklist before changing production configs:
Browse anthropic.com/news. Anthropic has never skipped a launch post for a production Sonnet tier. Rumor articles and X screenshots do not substitute.
curl https://api.anthropic.com/v1/models \
-H "x-api-key: $ANTHROPIC_API_KEY" \
-H "anthropic-version: 2023-06-01"
Look for a new Sonnet entry in the response. Until it appears here, third-party platforms showing claude-sonnet-5 in error logs are staging artifacts — the same class of signal that preceded Sonnet 4.6 in February.
Check Anthropic's pricing page. A model that is not priced is not generally available, regardless of what partner dashboards show.
If a new Sonnet does ship, compare against your frozen Sonnet 4.6 baseline on the same prompt set:
const models = ["claude-sonnet-4-6", "claude-sonnet-5"]; // swap ID when confirmed
for (const model of models) {
const res = await client.messages.create({
model,
max_tokens: 1024,
messages: [{ role: "user", content: evalPrompt }],
});
// log latency, token count, pass/fail against rubric
}
Do not migrate Claude Code or CI pipelines on slug sightings alone. Treat partner-platform leaks the way you would treat an unverified dependency version in package-lock.json — interesting, not actionable.
Use this when Sonnet 5 becomes real — not before.
| Step | Action |
|---|---|
| 1 | Export current Sonnet 4.6 eval results (SWE-bench subset, agent tasks, latency p95) |
| 2 | Document model ID strings in .env, CI configs, and Claude Code settings |
| 3 | Check Effort defaults — Sonnet 4.6 defaults to High; confirm new model behavior |
| 4 | Compare input/output pricing; update cost dashboards |
| 5 | Test tool-use and MCP integrations — see Claude Code MCP guide |
| 6 | Roll out canary traffic (5–10%) before full cutover |
| 7 | Keep Opus 4.8 as escape hatch for tasks Sonnet cannot handle |
If Fable 5 returns alongside Sonnet 5 with geographic restrictions, your routing logic may need three tiers: Fable (verified US), Opus (frontier fallback), Sonnet (global mid-tier). That is the scenario @kimmonismus flagged on X — structurally plausible, unconfirmed by Anthropic.
While Fable 5 sits suspended and Sonnet 5 remains rumor, developers are not waiting. GLM-5.2 topped BridgeBench Reasoning within 24 hours of the export ban. Kimi K2.7-Code shipped the same day as an open 1T-parameter coding model.
That context matters for Sonnet 5 rumors: Anthropic may accelerate Sonnet releases to retain developers who migrated during the Fable 5 gap — but acceleration via rumor is not acceleration via API. The competitive clock is running whether or not Sonnet 5 drops today.
For closed-API alternatives at lower cost, OpenRouter Fusion remains the most cited Fable-class substitute among ExplainX readers evaluating post-ban stacks.
Do not pause production work waiting for Sonnet 5.
claude-sonnet-4-6 todayIf you need Fable-class performance without Anthropic access, alternatives emerged within 48 hours of the ban — see GLM-5.2's post-ban benchmark results and OpenRouter Fusion as a lower-cost frontier option.
Claude Sonnet 5 rumors are louder on June 30, 2026 than at any point since February — but louder is not the same as confirmed. Partner slugs have misfired before. Anthropic's production Sonnet remains 4.6. Fable 5 remains suspended.
The productive move is to keep building on Sonnet 4.6, monitor official Anthropic channels, and have a migration checklist ready for the day — if it comes — when all three release signals align.
Last updated June 30, 2026. Model availability, slug sightings, and X speculation change quickly — verify against anthropic.com/news and the Models API before changing production integrations.