All articles
Company
4 min read
Introducing BLACKBOX07 v2.0
Agentic evals, voice routing and a completely rebuilt trace explorer. Here’s everything that’s new.
BLACKBOX07 Team
The biggest release since launch
Version 2.0 brings agentic evals that score multi-step tool use end to end, realtime voice routing under 300ms and a trace explorer that can diff any two runs side by side.
Existing projects upgrade automatically. Nothing breaks, and your current evals keep running.
What’s next
Self-hosted data planes are rolling out to Scale customers this quarter, followed by fine-tune hosting and a public model leaderboard built on real production evals.
Next article