Claude Capybara

Claude Capybara

$Capybara on Solana

Anthropic's leaked next-gen AI model. Bigger than Opus. Training complete. The codename that was never supposed to go public.

CA:
DfEueKFPabHT3DzywmKkXdPRns747QC2KrKSxuYxpump
Claude Capybara Banner

What is Claude Capybara?

On March 26, 2026, a sourcemap file was accidentally left in Anthropic's npm package for Claude Code. The entire source code was exposed. Inside, researchers found references to an unreleased model codenamed Capybara.

According to leaked draft blog posts, Capybara is a new tier of AI model above Opus, Anthropic's current flagship. The posts describe it as getting "dramatically higher scores on tests of software coding, academic reasoning, and cybersecurity, among others" compared to Claude Opus 4.6.

Two versions of the announcement were found. One calls it "Mythos," the other calls it "Capybara." Both use the same subtitle: "We have finished training a new AI model." Anthropic was deciding between names. Training is complete. The model exists.

What Else Was Leaked About Claude Capybara

🧠
Opus 4.7 + Sonnet 4.8
Migration code reveals at least two more model versions in development before Opus 5.
🤬
Frustration Telemetry
Anthropic tracks when users swear at Claude. Dedicated telemetry for user frustration and "keep going" prompts.
📝
Two Draft Blog Posts
Identical posts found. One says "Mythos," one says "Capybara." They were choosing the name when the leak happened.
🏗️
Above Opus Tier
Capybara is described as a new tier: larger and more capable than Opus, but also more expensive to run.

Claude Capybara: The Opus Successor

Claude Capybara in the Leaked Source Code

// claude-code/migrations/models.ts
const UNRELEASED_MODELS = {
"capybara": "claude-capybara-4.6",
"fennec": "claude-opus-4.6",
"opus-4-7": "claude-opus-4.7",
"sonnet-4-8": "claude-sonnet-4.8",
}
// WARNING: Do not expose unreleased model version numbers

Claude Capybara and Claude Mythos Sources