June 4 in X (former Twitter) appeared screenshot from the price list of the aggregator cc4 Groups: among the available models found a line claude-oceanus-v1-p. Price, context, latency, everything is in place. There is no official announcement from Anthropic. Let's see what that means.
In social networks, this model is discussed under the name Claude Mythos - Oceanus is the internal identifier of a particular checkpoint, which is now undergoing red team testing.
Where did the name come from and what is the screenshot?
In the interface of the cc4 Groups aggregator, which collects models from different providers in one API, a model with an identifier was lit up. claude-oceanus-v1-p.
The screenshot shows Input 16 $ / 1M tokens, Output 80 $ / 1M, Cached 1.6 $. Context window of 1M tokens, latency of 9.16 s, speed of 52 tokens per second. The status is partially red - the model is either not yet active publicly or in the process of rolling out.
suffix -p Similar to previous leaks, it means preview. This is what Claude Mythos Preview looked like before the official announcement.
Claude Oceanus and the line of models: where it fits
To understand scale, you need context. In March 2026, due to an error in the Anthropic CMS, drafts with a description of the Claude Mythos model were released to the public. The company confirmed the existence of the new development and called it a “qualitative leap” compared to everything before. In April, Mythos Preview was launched privately through Project Glasswing, only for cybersecurity organizations. The model never came out in public: too dangerous opportunities in terms of automation of cyber attacks.
On May 28, Anthropic released Claude Opus 4.8, an update to the flagship with improved coding and agent tasks. Opus 4.8 is inferior to Mythos, but “Mythos-level models are expected in the coming weeks.”
Claude Oceanus appears immediately after this announcement. A coincidence is unlikely.
What the price tag says $16/$80
Compare this to what we already know:
Claude Opus 4.8 (public): $5/$25 for 1M tokens
Claude Mythos Preview (closed, Project Glasswing only): $25/$125 for 1M tokens
Claude Oceanus v1-p (leaked): $16/$80 for 1M tokens
Oceanus is more expensive than the public Opus, but significantly cheaper than the closed Mythos. It’s more like a public version of Mythos-class – with some of the features, but without the sharpest cyber tools that Anthropic is not ready to give to everyone.
For comparison, GPT-5.6 is rumored to be preparing for release in the same time window. The price war between Anthropic and OpenAI is clearly going down, and $16/$80 fits in.
When's the announcement?
No direct dates. But there are two signals.
One: Anthropic publicly promised "Mythos-level models in the coming weeks" at the end of May.
Two: The model has already appeared in the price list of the aggregator, which means that the API is at least partially ready.
Both signals point to June 2026 as a realistic horizon.
Anthropic had no intention of making Mythos public at all. If Oceanus does come out for everyone, it’s a reversal of strategy. Either the company managed to develop rather rigid constraints, or it is a separate model with a different profile of capabilities.
What does this mean for those who use Claude at work?
If Oceanus is a public version of Mythos, the layout changes for several task categories.
So far, code audits and vulnerabilities have only been available to Project Glasswing partners. If Oceanus goes to an open API, these tools will be available to any developer – without a request or verification.
Agential tasks with a long horizon will have a more powerful foundation. Opus 4.8 is already good at multi-step scenarios, but Anthropic admits that Mythos is a different league.
The $16/$80 price tag does not look like a barrier. For professional use, this is a working range, especially considering caching.
So far, it's just a leak. Keep an eye on the official announcement.
What happened after the leak?
Developments have been rapid. Soon after the model appeared in the price list of the aggregator, it turned out that one of the participants of the red team resold access to the device. claude-oceanus-v1-p via a Chinese proxy service. Anthropic has suspended access for the entire test team and launched an internal investigation. On June 6, an identifier appeared in one of the AI trackers for a few minutes. Claude-Mythos-5 - and then it was deleted. What's behind him is still unknown.
If you only understand the terms AI-models – look at our glossaryIt's disassembled. RAG. token. context window and other basic concepts.
