Groq overcharging by 10x

Using GPT OSS 120B in Claude Code (via Claude Code router, then Cloudflare AI Gateway, then Groq).

Both the local Claude Code stats and Cloudflare’s AI gateway are showing roughly the same token usage.

However in Groq’s console, the tokens are shown around 10x what they are, leading to way higher cost than expected.

Anyone else encountered this?

Hi there! Sorry to hear that you’re encountering billing issues - please reach out to our support team at support@groq.com