Version 16. That’s not a typo. Oh My Pi is on version 16 while most coding agents are still in single digits. The latest release — v16.3.4 — shipped today with real improvements.
What’s New in v16.3
Baseten Provider Integration — Oh My Pi now supports Baseten as a model provider. Baseten is the inference platform that gives you fast, scalable access to open-source models. This joins an already deep provider list including OpenAI, Anthropic, Google, Groq, and local models.
Smarter Proactive Hard-Blocking — The AI usage system now exempts specific models from proactive hard-blocking. This means the agent knows which models are safe for which tasks and only hard-blocks when genuinely needed. Less friction, more flow.
Fixed Mnemonic Extraction — Mnemonic structured extraction now preserves empty results correctly. A subtle but important fix for workflows that rely on structured data extraction.
Usage CLI Reporting — The CLI reporting for usage was fixed and verified. You can now see accurate usage stats directly from the command line.
Why v16 Exists
Oh My Pi has been in continuous development longer than almost any other open-source coding agent. It was forked from Pi (by Mario Zechner) and has evolved through 16 major versions while maintaining backward compatibility and growing its community to 16K+ stars.
The version number reflects maturity. Oh My Pi has been production-tested at scale in ways that newer agents haven’t.
What This Means
Oh My Pi is the dark horse of coding agents. It doesn’t get the same attention as Hermes or the same ecosystem as OpenCode, but its feature set — LSP integration, hash-anchored edits, browser automation, subagents — is among the most complete in the space. Version 16.3 proves it’s still iterating fast.