Up to 5× faster RL training steps, on the same GPU budgets.
Our internal auto-tuner optimizes the full RL loop: trainin | Hanami
Up to 5× faster RL training steps, on the same GPU budgets.
Our internal auto-tuner optimizes the full RL loop: training, inference, and environment execution. It profiles each job, forecasts promising configuration changes, tests them in real runs, and follows the bottleneck as it moves.
Across 210 runs on NVIDIA B300 GPUs, we tested four dense and MoE models on math and agentic coding workloads. All eight jobs improved: 1.42-4.93x faster steps, with a median speed-up of about 3.5x.
One revealing result: nearly doubling inference-engine throughput in a benchmark didn’t shorten the RL step. Sandbox execution absorbed the gain.
That’s why we measure the complete job, alongside numerical checks and longer reward comparisons. GPU counts stayed fixed; some configurations added sandbox hosts.
We’ve shared the system, results, and five engineering lessons, including what didn’t work, in the article below:
Retweeted by @nebiusai from @nebiustf(Historical earnings: For fiscal 2027 Q2 (period ended 2026-07-26), NVIDIA reported basic EPS of 4.87, diluted EPS of 4.85, net income of USD 118.010 billion, and revenue of USD 177.837 billion.
Consensus expectations: For fiscal 2027 Q3 (event date 2026-11-17), consensus expects EPS of 2.4733 and revenue of USD 108.977 billion.)