Anthropic may be preparing an unexpected new Claude release, and this time the jump could be bigger than a routine model refresh.
Multiple reports circulating in the AI developer community claim that Anthropic is internally testing a model identified as Claude Opus 5.5, with the internal codename claude-wafer-eap.
The most eye-catching part of the rumor is the alleged launch timing. The model could arrive as soon as Tuesday, September 22, although Anthropic has not announced a release date.
Even more interestingly, leaked pricing reportedly puts Opus 5.5 below the current Claude Opus 5 on both input and output tokens.
There are also claims of significant improvements in coding, long-running agent tasks and 3D generation.
For now, however, all of this remains unofficial.
Claude Opus 5.5 could arrive on Tuesday
The current rumor points to Tuesday, September 22, as the possible launch date for Claude Opus 5.5.
Developer and community posts claim Anthropic is already running closed or stealth testing of the model under the codename claude-wafer-eap. Multiple reports have repeated the Tuesday timing, although the exact release schedule remains unconfirmed.
There is an important distinction here.
Anthropic has not publicly announced Claude Opus 5.5.
The company’s official model documentation currently lists Claude Opus 5 as an active model, while Anthropic’s latest newsroom announcements include Claude Fable 5.1 and Claude Mythos 5.1 rather than an Opus 5.5 launch.
That means the reported Tuesday launch should be treated as a leak, not a confirmed release date.
The rumored pricing could be the bigger story
If the leaked information is accurate, Anthropic may be preparing something particularly interesting for developers.
The reported API pricing for Claude Opus 5.5 is:
| Model | Input / 1M tokens | Output / 1M tokens |
|---|---|---|
| Claude Opus 5 | $5 | $25 |
| Claude Opus 5.5 rumored | $4 | $20 |
Anthropic’s official documentation confirms that Claude Opus 5 currently costs $5 per million input tokens and $25 per million output tokens.
The alleged Opus 5.5 pricing would therefore represent a 20% reduction on both input and output token costs.
That’s notable because the rumored model would supposedly be a newer and more capable version rather than simply a cheaper alternative.
The leaked figures also reportedly include much lower prompt-cache read pricing, although those numbers are not officially confirmed.
Why cheaper pricing matters
For developers building AI agents, token pricing can quickly become a major part of operating costs.
An agent may repeatedly:
- Read large codebases
- Analyze documents
- Call external tools
- Generate large amounts of code
- Run multiple reasoning steps
- Maintain long conversations
- Repeat tasks while checking its own work
A 20% reduction in both input and output prices could therefore make a meaningful difference at scale.
The actual savings would depend on how many tokens a workload consumes, how much cached context is used and how many model calls an application makes.
Why the jump to 5.5 is attracting attention
Another reason the rumor is generating discussion is the version number.
Claude Opus 5 launched in July 2026. Anthropic’s official announcement described it as a major model focused on reasoning, coding and knowledge work.
A move directly to 5.5 would suggest something more substantial than a minor point release—at least from a naming perspective.
That doesn’t automatically mean Opus 5.5 will deliver a massive capability jump.
Model version numbers are ultimately a product decision made by the company, and Anthropic has not explained what the alleged 5.5 designation would represent.
Still, the unusual numbering has fueled speculation that Anthropic could be preparing a more substantial update.
3D generation is one of the biggest rumored improvements
Some of the most interesting claims surrounding the leaked model involve 3D generation and coding.
Community testers have reportedly used the model to create interactive 3D experiences, including vehicle models and browser-based 3D applications using technologies such as Three.js.
These demonstrations are particularly interesting because Claude’s recent evolution has increasingly focused on models acting as coding and engineering agents rather than simply answering questions.
However, screenshots and demonstrations posted by individual testers shouldn’t be treated as independent benchmark results.
There is currently no official Anthropic benchmark data for Claude Opus 5.5.
So claims that the model is dramatically better at 3D generation—or that it beats another unreleased or rumored AI model—remain claims rather than established facts.
Claude is already moving toward longer-running AI agents
The rumored improvements would fit with a broader direction already visible in Anthropic’s recent models.
Claude Opus 4.8, for example, was introduced with stronger agentic capabilities and support for longer-running workflows. Anthropic highlighted dynamic workflows that allow Claude Code to plan work, use parallel subagents and verify outputs before returning results.
Claude Opus 5 subsequently continued that push.
Anthropic describes Opus 5 as a model aimed at deep reasoning, coding and knowledge work, with a 1-million-token context window and up to 128,000 output tokens according to its documentation.
That makes the rumored focus on coding, planning and 3D development plausible as an area Anthropic could continue improving.
But again, plausible isn’t the same as confirmed.
The timing would put Anthropic back in the AI spotlight
A potential Opus 5.5 launch would also arrive during an unusually competitive period for frontier AI models.
OpenAI and Google are both expected to continue pushing their next-generation models, while Anthropic has been rapidly expanding Claude’s capabilities across coding, enterprise workflows and scientific applications.
Anthropic’s latest official announcements show how quickly the company is expanding its model lineup. Claude Fable 5.1 and Claude Mythos 5.1 were announced on September 1, just weeks after the Opus 5 launch.
A new Opus model only a couple of months after Opus 5 would represent another acceleration in that release cycle.
What we know versus what is still a rumor
It is worth separating the confirmed information from the speculation.
Confirmed
- Claude Opus 5 is an official Anthropic model.
- Opus 5 launched on July 24, 2026.
- Its official API price is $5 per million input tokens and $25 per million output tokens.
- Anthropic currently lists Opus 5 as an active model.
- Anthropic has officially announced Claude Fable 5.1 and Mythos 5.1.
Not confirmed
- Claude Opus 5.5 exists as a publicly released model.
claude-wafer-eapis the final internal codename for the model.- The model will launch on September 22.
- The rumored $4/$20 API pricing is final.
- The model delivers a major improvement in 3D generation.
- Opus 5.5 beats competing frontier models.
Those details currently come from leaks and community reports rather than an official Anthropic announcement.
Could Opus 5.5 really be cheaper?
That’s perhaps the most interesting part of the leak.
The alleged price reduction would make the model cheaper to run than Opus 5 while potentially offering improved capabilities.
For developers, that combination is much more important than the version number itself.
A model that is 10–20% better but significantly more expensive can be difficult to deploy broadly. A model that is more capable and cheaper per token is much easier to justify for large-scale applications.
Of course, the final economics could look different once Anthropic publishes official pricing.
Subscription plans such as Claude Pro, Max or enterprise offerings may also have different limits and pricing structures, so API pricing should not automatically be interpreted as a change to consumer subscription costs.
What happens next?
If the Tuesday rumor is accurate, we may not have to wait long for answers.
The most important things to watch will be:
- The official model name
- API availability
- Final input and output pricing
- Context-window limits
- Maximum output size
- Coding and agentic performance
- 3D and multimodal capabilities
- Claude Code availability
- Consumer plan availability
- Independent benchmark results
Those details will tell us far more than leaked screenshots or individual demonstrations.
Anthropic has not confirmed Claude Opus 5.5 yet
For now, Claude Opus 5.5 remains a leaked and unconfirmed model.
The combination of a possible Tuesday launch, a jump to the 5.5 name and rumored lower API pricing is enough to make the story interesting, but none of those details should be treated as official until Anthropic publishes them.
Anthropic’s current documentation still identifies Claude Opus 5 as its active Opus model, with official pricing of $5 per million input tokens and $25 per million output tokens.
If the leaked $4/$20 pricing turns out to be real, however, the announcement could be particularly significant for developers running AI agents at scale.
And if the reported improvements in coding, long-horizon tasks and 3D generation also hold up under independent testing, Claude Opus 5.5 could quickly become one of the most closely watched AI model releases of September.
For now, the safest takeaway is simple: the leak is intriguing, the pricing claims are notable, but Anthropic still has the final word.
For continuing updates on OpenAI, ChatGPT, Codex, GPT models, and AI industry developments, bookmark:
