Anthropic may be preparing another major update to its Claude model lineup, with reports pointing to a model identified as Claude Opus 5.5 being tested internally ahead of a possible September 22 release. However, Anthropic has not publicly announced Opus 5.5, published a model card, listed an API identifier, or confirmed the reported launch date.
The claims currently circulating around Claude Opus 5.5 include an alleged internal codename, claude-wafer-eap, along with proposed pricing and performance characteristics. Those details originate from leaks and social-media reports rather than Anthropic’s official documentation, so they should be treated as unverified until the company provides confirmation.
Anthropic’s official model documentation currently lists Claude Opus 5 as an active model. The company’s most recent fifth-generation releases include Claude Fable 5.1 and Claude Mythos 5.1, both introduced in September. There is currently no official Opus 5.5 entry in Anthropic’s model documentation or system-card archive.
Quick Summary
- Claude Opus 5.5 is reportedly being tested internally by Anthropic.
- The reported codename is
claude-wafer-eap. - Reports suggest a possible September 22 release, but Anthropic has not officially confirmed it.
- Leaked claims point to improvements in coding, frontend development and 3D-related workflows.
- Reported pricing and context-window figures remain unverified.
- Claims that Opus 5.5 definitively outperforms other frontier models should not be presented as established facts.
- The article separates verified Anthropic information from unconfirmed leak claims.
What the Claude Opus 5.5 Leak Claims?
The reported Opus 5.5 story centers on the possibility that Anthropic is testing a new Opus model before making it publicly available.
One report attributes the claude-wafer-eap codename to a leak and says the model could be released on Tuesday, September 22. The same reporting notes that there is currently no publicly available model card, API identifier, official pricing page or independent benchmark that verifies the leaked specifications.
A separate analysis of the claims reports that an alleged pricing sheet lists $4 per million input tokens and $20 per million output tokens, alongside cache pricing of $0.20 per million tokens for reads and $5 per million for writes. Those figures have not been published by Anthropic and therefore should not be treated as the confirmed pricing for Claude Opus 5.5.
The same applies to the reported 872K-token context window. Anthropic has not published an official specification confirming that figure.
Anthropic’s Current Opus Model
The most important verified reference point is Claude Opus 5.
Anthropic introduced Claude Opus 5 on July 24, 2026, describing it as a model designed for coding, knowledge work and longer-running agentic tasks. The company says Opus 5 is available across its platforms and through the Claude API.
Anthropic’s official launch information lists Opus 5 at $5 per million input tokens and $25 per million output tokens. It also supports a 1-million-token context window and up to 128,000 output tokens under the standard configuration described by Anthropic.
The model is positioned around software engineering, knowledge work, computer use and scientific research. Anthropic’s own evaluations report strong results across several internal and external benchmarks, although those results are company-reported and should be interpreted within the specific benchmark conditions.
That provides the current verified baseline against which any future Opus 5.5 release would need to be evaluated.
Why the Reported Opus 5.5 Specifications Matter?
If the leaked pricing eventually proves accurate, the reported $4 input and $20 output rates would represent a reduction from Opus 5’s official $5/$25 pricing. That would make pricing one of the potentially significant changes between the current Opus model and the rumored successor.
However, it is too early to describe the alleged pricing as a confirmed reduction. Anthropic can change pricing, context limits, availability and other specifications before a model reaches public release.
The reported context window also deserves scrutiny. Claude Opus 5 officially supports a 1-million-token context window, so an 872K context window would actually be smaller than the current Opus 5 specification. Whether that figure is accurate, temporary, or associated with a particular testing environment is unknown.
What About 3D, Frontend and Coding Performance?
The supplied report claims that the alleged Opus 5.5 performs particularly well on 3D and frontend development, with improvements in visual detail and polished outputs.
There is some relevant context here, but not independent confirmation of those Opus 5.5 claims.
Anthropic itself highlighted Claude Opus 5’s ability to produce visual outputs and reported examples involving 3D reconstruction and interactive software. The company’s launch material also includes feedback from developers describing improvements in frontend work, animations, games and 3D tasks.
Those examples demonstrate that 3D and frontend development are already areas of interest for the Opus family. They do not, however, establish that an unreleased Opus 5.5 has the specific improvements described in the leak.
No Verified Evidence That Opus 5.5 Beats GPT-6 Sol
One of the strongest claims in the supplied report is that Opus 5.5 would outperform GPT-6 Sol.
There is currently no reliable evidence supporting that as a factual statement.
No official Anthropic benchmark establishes such a comparison, and an unreleased model cannot meaningfully be declared the winner of a broad model comparison without reproducible testing under defined conditions.
For that reason, the claim should not appear in the article as a factual performance result. Once a model becomes publicly available, comparisons could instead examine specific benchmarks, coding tasks, reasoning evaluations, latency, pricing and real-world workloads.
What Would Confirm Claude Opus 5.5?
Several pieces of evidence would quickly distinguish a genuine launch from an unconfirmed leak.
The strongest signals would be an Anthropic announcement, a new model entry in Anthropic’s documentation, an official API identifier, a model card or system card, and an official pricing update.
Anthropic’s system-card archive is particularly useful because the company publishes capability and safety documentation for its major Claude models. The current archive includes Claude Opus 5 and the September 2026 Fable 5.1 and Mythos 5.1 releases, but not Opus 5.5.
A publicly accessible API model identifier would also provide substantially stronger evidence than an internal codename circulating through social media.
What We Know About Claude Opus 5.5 Right Now?
For now, the evidence supports a narrower conclusion: Claude Opus 5.5 is a reported upcoming model, not a confirmed Anthropic release.
The reported claude-wafer-eap codename, September 22 release date, $4/$20 pricing, 872K context window and performance claims remain unverified. Recent independent reporting has similarly identified the leak as single-source or otherwise unconfirmed.
Anthropic’s verified product lineup remains the appropriate reference point until the company publishes further information. If Opus 5.5 is officially released, its final pricing, context window, capabilities and benchmark results should be evaluated against the specifications of Claude Opus 5 rather than against leaked figures.
Editorial Note: Anthropic has not officially announced Claude Opus 5.5. Details about its release date, pricing, context window, codename, and performance remain unverified and are presented in this article as reported claims, not confirmed specifications.
Also Read –
Claude Opus Fast Mode: High-Speed Opus AI Execution
Source
Anthropic – Introducing Claude Opus 5
Anthropic – Model Deprecations and Current Model Status
Anthropic – Model System Cards
Anthropic – Introducing Claude Fable 5.1 and Claude Mythos 5.1
OrcaRouter – Claude Opus 5.5 Leak Analysis
CellCog – Claude Opus 5.5 Release-Date and Pricing Analysis
