Claude Opus 4.6 did not get a verified formal shutdown announcement. The clearest primary-source evidence points to a newer model release, changing product surfaces, and documentation that lagged behind what some users were seeing in the app. Here, a formal deprecation means Anthropic publishes a retirement or sunset signal in model docs, changelogs, or migration notices, while a product-surface shift means the model may disappear from a UI, plan, or region without being fully retired. Anthropic announced Claude Opus 4.7 on April 16, 2026. On the same day, Anthropic’s official Opus page still described the Opus lineup rather than posting a deprecation notice for Claude Opus 4.6, which is a different signal from a hard retirement. Anthropic’s official Claude Opus model page says Opus 4.7 is the current flagship model. Anthropic reports 13% better performance than Opus 4.6 on its 93-task coding benchmark, plus availability across Pro, Max, Team, Enterprise, the Claude API, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry. That is a rollout. It is not, by itself, a shutdown notice for Claude Opus 4.6. Launch pages normally do a specific job: they announce the new default, list where the new model is available, and give benchmark and pricing context. Retirement usually shows up somewhere else, with deprecation labels, sunset dates, migration guidance, or explicit changelog language telling users what is going away and when. Those retirement signals were not present in the Anthropic sources reviewed here. The same product page gives pricing context: $5 per million input tokens and $25 per million output tokens for Opus 4.7, with US-only inference at 1.1x pricing, according to Anthropic. Those are launch details for a replacement-tier model, not language saying older access was terminated. Anthropic’s newsroom lines up with that timing. It shows “Introducing Claude Opus 4.7” on April 16, 2026, followed by other product updates the next day. There is no matching newsroom post saying Anthropic discontinued Claude Opus 4.6. A short version: Claim Verification Source Opus 4.7 launched on April 16, 2026 Verified Anthropic Opus page; Anthropic Newsroom Opus 4.7 is available across paid plans and major cloud channels Verified Anthropic Opus page Anthropic formally deprecated Claude Opus 4.6 last week Unverified No primary-source announcement found Some users may have seen Claude Opus 4.6 disappear from product surfaces Plausible inference Consistent with new-model rollout plus no formal deprecation notice The verified evidence in this case is narrow: Anthropic launched a new flagship model, and Anthropic’s public materials still referenced Claude Opus 4.6 in other contexts. That combination is consistent with UI, entitlement, or region-level changes, but it does not prove any specific app behavior without a product bulletin or support documentation saying so. That distinction matters because availability is not one switch. A model can be visible in one place and absent in another. A simple process view: Surface What it controls What the reviewed sources show Product page/docs Official lineup, benchmarks, pricing Verified: Opus 4.7 presented as current flagship Newsroom/changelog Launch timing and formal announcements Verified: Opus 4.7 launch announcement; no 4.6 retirement post found Chat UI Which models appear in selectors Not verified in primary sources reviewed here Plan tier entitlements Which subscribers can access which models Verified for Opus 4.7 availability; not verified for 4.6 removal Region/cloud channel Where a model can be used Verified for Opus 4.7 multi-channel availability; not verified for 4.6 withdrawal Anthropic’s own recent pages cut against the idea that Claude Opus 4.6 was abruptly removed for safety or policy reasons. The Responsible Scaling Policy page states that Claude Opus 4.6 does not cross the threshold discussed there. That is an official current statement about its status. There is also fresh technical reporting that still references the model as an active subject of analysis. In Anthropic’s BrowseComp eval-awareness post, the company reports 9 contaminated examples out of 1,266 BrowseComp problems in one evaluation setup involving Opus 4.6. That is recent technical treatment of the model, not a retirement notice. The catch: “I can’t select it anymore” and “it was deprecated” are not the same claim. For users, the main verified change is that Opus 4.7 becomes the visible flagship in Anthropic’s public materials. Anthropic says it is stronger on coding, vision, and multi-step tasks, and the company’s headline number is a 13% coding benchmark gain over Opus 4.6. For teams, the practical issue is less about headlines and more about whether their workflows are tied to a specific model version. Area Verified fact Practical precaution Model selection Anthropic’s Opus page presents 4.7 as the flagship model Verify whether your workflow uses an explicit model ID or a moving default Access channels Anthropic says Opus 4.7 is available in the Claude API and major cloud platforms Check availability separately in the Claude app, API, Bedrock, Vertex AI, and Foundry Performance expectations Anthropic reports a 13% gain over Opus 4.6 on its 93-task coding benchmark Re-run internal evals before assuming a newer default will behave the same in production Pricing Anthropic lists Opus 4.7 pricing at $5/million input tokens and $25/million output tokens Confirm cost assumptions if your traffic shifts to the newer flagship That is the operator gotcha. If a workflow depends on a particular model and nobody has pinned or verified it, a quiet availability change can look like a quality regression, latency change, or prompt drift when the underlying issue is simpler: the request path changed. Anthropic’s product page also says Opus 4.7 is available through multiple managed channels. That matters because Anthropic model availability is often channel-specific. A model can appear in the direct API before it appears in every app surface, or stay visible in one cloud channel while being less prominent elsewhere. Formal deprecations are typically documented in changelogs, model docs, or migration notices; anecdotal availability reports are weaker evidence on their own. Use this checklist before treating Claude Opus 4.6 as retired: Signal What it usually means How strong it is Official deprecation post or changelog Planned retirement or removal date Strong API docs mark model as deprecated/sunset Access likely ending Strong Migration notice to move off a model Retirement is being operationalized Strong Newsroom launch of newer model only Rollout, not necessarily retirement Medium UI selector no longer shows model Surface or plan change Weak on its own Support anecdote or social post Possible early signal, needs confirmation Weak Anthropic has given strong evidence for the new model rollout. It has not, in the sources reviewed here, given strong evidence for a blanket Claude Opus 4.6 shutdown. That does not mean access is unchanged. It means the verified statement is narrower: availability may have shifted faster than documentation. Teams should treat the current state as a rollout with incomplete visibility, and check official model docs, changelogs, and channel-specific availability before assuming deprecation. Claude Opus 4.6 was not accompanied by a verified formal deprecation announcement in the primary sources reviewed here. Anthropic officially launched Claude Opus 4.7 on April 16, 2026, with a reported 13% gain over Opus 4.6 on its 93-task coding benchmark. Anthropic’s current public materials still reference Claude Opus 4.6 in both policy and technical evaluation pages. A model disappearing from a UI or plan tier can be a rollout or access change, not proof of retirement. Teams using Claude in production should check model IDs, changelogs, and channel-specific availability before interpreting behavior changes. Claude Opus model page — Anthropic’s official Opus lineup page with Opus 4.7 launch details, pricing, and availability. Anthropic Newsroom — Official announcement timeline showing when Opus 4.7 was introduced. Anthropic Responsible Scaling Policy — Anthropic’s policy page stating that Claude Opus 4.6 does not cross the threshold discussed there. Anthropic BrowseComp eval-awareness post — Engineering post with recent Opus 4.6 evaluation details, including contamination findings. Originally published on novaknown.com