Extrapolate

Scaling RSI

RL training environments for long-horizon SWE and ML.

Work with us

300+ environments ready for training

We’re building training environments for recursive self-improvement: an agent’s ability to help build the next version of itself.

We believe high-quality RSI data is the bottleneck. Our environments cover improving sample efficiency and model architectures, debugging training runs, and making open-ended research decisions.

We have partnered with several frontier labs and are committed to increasing sovereign intelligence worldwide.

Results

SemiAnalysis GDPx RSI results. Mean reward on a zero to one scale, mean cost per task in US dollars, and average output tokens in millions. Rows are sorted by descending reward. Missing public prices are shown as a dash.
Model cohortMean rewardMean cost per taskAvg. output tokens (M)
GPT-6 Astra0.2489$129.280.602
Gemini 4 Carbon0.1906$11.950.439
Claude Opus 5.50.1755$29.590.368
GLM 5.30.0712$5.270.116

4 models shown.

View on GDPx ↗

Our team comes from

Work with us

Tell us about your research or training needs.

research@extrapolate-ai.com