News

Releases & Models

RUMOR: Testers say they spotted an unannounced Claude inside Claude Code, calling it "Claude Next"

A tester with a track record of spotting unreleased models says he found, inside Claude Code, a model that makes Opus 5.5 look primitive. A second observer reports a similar feeling but bets on a different name. Anthropic has not commented.

Ilustracao editorial sobre o rumor do modelo Claude Next dentro do Claude Code

RUMOR

A tester with a track record of spotting unreleased Anthropic models before the company says anything claims to have found, inside Claude Code, a model he is calling Claude Next. According to the report, it makes the current flagship, Claude Opus 5.5, look primitive, and produced a fully coded animation from just a couple of prompts. A second observer with a large X following says his own traffic was being routed somewhere unusual and guesses it could be a fresh Opus checkpoint or even an Opus 5.6. Anthropic has said nothing so far.

What the tester says he saw

Claude Code is Anthropic's command line coding agent, used today by millions of developers who pick which model runs behind it. On October 11, the tester known as Chetaslua posted that his requests inside Claude Code were being answered by a model other than the Opus 5.5 he had manually selected. He called the result state of the art, said it would make current image and video generation tools obsolete within a month, and promised more tests soon. None of those posts included a model identifier, a documentation screenshot, a system log or any data a third party could verify, it is a usage report, not technical evidence.

A second account backs the mood, not the name

Hours later, another influential observer known as Chubby said his own Claude Code answers also seemed to come from somewhere else and looked really good, floating the idea that Anthropic might be routing some traffic to a new Opus 5.5 checkpoint or something like Opus 5.6. That is a second tester reporting the same sense that something changed, which gives the rumor some weight, but he misses exactly the part that matters most, it does not match the name the first tester used. Two reports agreeing something shifted while disagreeing on what shifted is still a rumor, not a confirmation.

Why Claude Next does not fit how Anthropic names models

Anthropic draws from a small, closed vocabulary of evocative names, Mythos, Fable, Opus, Sonnet and Haiku, always followed by a version number, as in claude-opus-5-5 or claude-fable-5-1. Next fits none of those families and carries no version number at all. There is a duller, more likely explanation, Claude Code ships on npm under two channels, a stable one and one literally called next, used to test the client itself before promoting it to stable. In other words, next is already a word floating around this ecosystem tied to the tool, not to the brain answering inside it, which makes it easy to mistake a channel label for a model name.

What is already confirmed practice, and why that proves nothing

Anthropic does test unreleased frontier models with small groups of selected users before deciding whether they ship publicly. Claude Mythos Preview is a documented example from the company itself, which even states it does not plan to release that model under that name. In October, Anthropic also expanded its verification program for cybersecurity researchers, explicitly saying new models will be folded into that restricted access as they appear. That shows the mechanism behind the rumor is plausible and has precedent, but it proves nothing about whether this specific Claude Next exists or will ever become a product.

MaxAssistant's read

This is backroom chatter, not a launch in progress. The shape of the claim is exactly what a well told leak should look like, a tester reports early access with no date, no version and no identifier, and says more tests are coming. It is worth watching because Anthropic has been shipping on short cycles since Fable 5.1 and has plenty of competitive reason to prepare something bigger after GPT 6.1 Sol and Gemini 4 Argon. But the real turning point is not another slick video made by an unidentified model, it is a model identifier showing up in Anthropic's own model table, in a Claude Code changelog or in a system card. Until that happens, Claude Next is a name in a post, and the four models Anthropic actually sells, Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5, remain the only ones worth your engineering time.

Sources

OrcaRouter: Claude Next leak, an unreleased Claude in Claude Code; TechCityAuthority: Claude Next could be Opus 5.6 as more testers report a new model.