Claude Capybara Release Date: When Will Anthropic Launch Mythos?
Anthropic has not announced an official release date for Claude Capybara. The model is currently in restricted early testing with a small group of cybersecurity defense organizations. Based on available evidence — the company’s IPO timeline, leaked internal documents, and release patterns — the most likely window for general availability is October 2026. See our the March 2026 leak for a deeper look.
Here is everything we know about when Capybara will become available, who gets access first, and what determines the timeline.
Current Status: March 2026
As of late March 2026, Capybara’s release status is defined by three confirmed facts from Anthropic:
In testing with early access customers. An Anthropic spokesperson confirmed the company is “working with a small group of early access customers to test the model.” These are not regular API users — they are specifically selected organizations with cybersecurity defense use cases.
Not ready for general release. Anthropic stated the model is “expensive to run and not yet ready for general release.” This signals both technical constraints (inference costs, infrastructure scaling) and policy constraints (safety evaluations incomplete).
Timeline determined by safety, not schedule. The company has indicated that rollout timing will be “determined by safety evaluation outcomes rather than a fixed commercial schedule.” This means no date is locked — the release could accelerate or delay depending on evaluation results.
The IPO Connection
The strongest signal for Capybara’s release timing comes from Anthropic’s potential IPO.
What Bloomberg Reported
On March 26, 2026 — the same day as the Capybara data leak — Bloomberg and The Information reported that Anthropic is considering an IPO as early as October 2026. The company’s valuation is estimated at $60 billion or more, making it one of the largest AI company IPOs in history.
Why IPO Timing Matters
Launching a breakthrough model before going public creates maximum market excitement and validates the company’s valuation. Every major tech IPO in the last decade has been preceded by a flagship product launch that demonstrates the company’s technical leadership.
For Anthropic, Capybara is that product. The Capybara tier represents capabilities that no competitor can match (according to leaked benchmarks), and a public launch before the IPO would provide concrete evidence of Anthropic’s technical moat to potential investors.
The October 2026 Theory
If Anthropic files for an IPO in October 2026, they would likely launch Capybara 2-4 weeks before the filing. This places the most probable release window in September or early October 2026. The launch would generate press coverage, developer adoption metrics, and revenue projections that strengthen the S-1 filing.
Predicted Release Timeline
Based on all available information, here is the most likely progression:
| Period | Expected Development |
|---|---|
| April-May 2026 | Expanded early access to additional enterprise customers |
| June-July 2026 | Limited beta or waitlist for qualified organizations |
| August-September 2026 | Safety evaluations completed, pricing announced |
| October 2026 | General availability via API (aligned with IPO) |
What Could Accelerate the Timeline
Competitive pressure from OpenAI or Google could force Anthropic’s hand. If a competitor releases a model with comparable cybersecurity capabilities, the strategic advantage of controlled release diminishes.
Successful safety evaluations could clear the path for earlier release. If the early access customers demonstrate that the model can be deployed safely, Anthropic may expand access ahead of schedule.
What Could Delay It
Safety concerns are the primary risk. If early testing reveals unexpected misuse vectors or the model’s cybersecurity capabilities prove more dangerous than internal assessments suggested, Anthropic will delay rather than release a model they consider unsafe.
Technical challenges with inference costs could also delay release. Anthropic confirmed the model is “expensive to run,” and scaling infrastructure for general availability at acceptable latency and cost may require additional engineering time.
Who Gets Access First
Anthropic has been explicit about the access priority order, driven by the model’s cybersecurity implications.
Phase 1: Cyber Defense Organizations (Now)
The first group to receive Capybara access consists of cybersecurity defense organizations — companies and government agencies whose primary mission is protecting systems from attack. Anthropic’s stated goal is giving defenders a “head start in improving the robustness of their codebases” before the model becomes broadly available.
Phase 2: Enterprise Early Access (Expected Q2-Q3 2026)
Large enterprise customers with existing Anthropic relationships will likely receive access next. This phase would include companies using Claude for internal development, research organizations, and major API customers. Pricing discussions and SLA negotiations would happen during this phase.
Phase 3: General API Access (Expected Q4 2026)
Full API availability for all developers. This is when individual developers, startups, and smaller organizations can access the Capybara tier through the standard Claude API. No waitlist, no special approval — just a model parameter change.
Phase 4: Consumer Products (Unknown)
Whether Capybara will be available through Claude.ai’s consumer interface (Max, Team, or Enterprise plans) is unknown. Given the pricing expectations (2-5x Opus), consumer-facing Capybara access may be limited to premium subscription tiers or usage-capped.
API Preparation
When Capybara becomes available, the transition will be simple for developers already using the Claude API.
What You Can Do Now
Build your applications using Claude Opus 4.6 or Sonnet 4.6. The Claude API uses a unified interface — when Capybara launches, switching requires only changing the model parameter in your API calls. No code restructuring, no SDK changes.
Test your prompts and workflows with Opus 4.6 to establish baselines. When Capybara becomes available, you can compare performance directly and decide which tasks justify the premium pricing.
What the Model ID Will Look Like
Based on Anthropic’s naming convention, the Capybara model ID will likely follow the pattern claude-capybara-mythos-YYYYMMDD or similar. The exact ID will be published in Anthropic’s documentation when the model launches.
Comparison to GPT-5 Release Strategy
OpenAI’s approach to GPT-5 offers an instructive contrast.
GPT-5: Fast Launch, Mixed Reception
OpenAI released GPT-5 broadly in August 2025, prioritizing speed to market. The model was available through the API and ChatGPT within weeks. The result was widespread availability but also widespread disappointment — the model fell short of pre-release promises according to many users and reviewers.
Capybara: Controlled Launch, Unknown Reception
Anthropic is taking the opposite approach: no marketing hype, no benchmark promises (the leaked ones were unintentional), and a phased rollout driven by safety evaluation rather than commercial schedule. The risk is that by the time Capybara launches, competitors may have closed the capability gap. The benefit is that if it delivers on the leaked benchmarks, the reception will be earned rather than marketed.
Questions About Claude Capybara Release Date
When will Claude Capybara be released?
No official date has been announced. Based on Anthropic’s potential October 2026 IPO timeline and current early testing status, general API availability is most likely in late 2026, potentially September or October.
Is Claude Capybara available now?
No. It is in restricted testing with a small group of cybersecurity defense organizations. There is no public API access, no waitlist, and no consumer availability.
How can I get early access to Capybara?
Anthropic has not published an application process. Current early access is limited to cybersecurity defense organizations selected by Anthropic. Enterprise customers with existing Anthropic relationships may receive priority for the next access phase.
Will Capybara be available on Claude.ai?
Unknown. Given expected premium pricing (2-5x Opus), consumer access may be limited to high-tier subscription plans. API access for developers is more likely for the initial release.
Could the release be delayed?
Yes. Anthropic has stated that the timeline is determined by safety evaluation outcomes. If testing reveals concerns about misuse potential, the release will be delayed. Competitive pressure could accelerate it.
How much will Capybara cost when released?
No official pricing. Estimates range from 2-3x to 4-5x current Opus pricing ($15/M input, $75/M output). Batch API discounts of 50% will likely apply for non-real-time workloads.