OpenCode inference tops 6 trillion daily tokens
OpenCode Inference now handles over 6 trillion tokens per day, roughly three times the volume of the OpenCode Go dataset. The team is currently cleaning this telemetry data to ensure accuracy before fully integrating it into their public model usage stats.
Dax Raad's announcement highlights the massive scale at which open-source AI agent tooling is being adopted, showing that self-hosted or user-provided key setups represent the vast majority of usage over managed subscriptions.
- –**Massive Inference Scale:** Handling over 6 trillion tokens daily positions OpenCode as a highly active player in the developer agent ecosystem.
- –**The Self-Hosted Majority:** With total inference at 3x the OpenCode Go dataset, most users prefer configuring their own API keys rather than using the managed subscription endpoints.
- –**Data Cleanup Challenges:** Dax notes the need to filter out quirks and anomalies from the userbase to make the public telemetry dataset truly representative and meaningful.
DISCOVERED
50d ago
2026-06-11
PUBLISHED
50d ago
2026-06-11
RELEVANCE
AUTHOR
thdxr