The model is the lab. The lab never sleeps.
Benchmarks, environments and the open-source stack behind every PI training loop.
Browse benchmarks
Benchmarks How we measure self-improvement: cycle-over-cycle gains, time to first improvement, and regression gates. Open → 02
Environment Hub Browse and compose 2,500+ open-source RL environments — or synthesize private ones from production traces. Open → 03
Open-source stack Verifiers, Prime-RL and Sandboxes — the open components that run every training loop. Open → FIG. 01 / EFFICIENCY
3.2×
Avg. improvement per cycle
FIG. 02 / VELOCITY
48 h
First improvement deployed
FIG. 03 / SCALE
100B+
Parameters on-stack
FIG. 04 / LABOR
0
Reward functions written
Self-generating eval suites- Auto-instrumented agents
- RL across 2,500+ environments
Serverless model inference- Real-time training observability
Production data flywheel
[ Lab access ]
Run your first training cycle this week.
Engineers, at full stride. Point PI at your checkpoint — first improvement in about 48 hours.