Bespoke Labs post-trains Nemotron-3.5-Super-VL into an ML research agent, scoring 35%

Bespoke Labs says it post-trained NVIDIA's Nemotron-3.5-Super-VL into an autonomous ML researcher, raising its score on AutoResearchExam from 28% to 35%.

Bespoke Labs says the result matches Qwen 3.8 Max, which it describes as having 20 times the total parameters. The announcement was shared on X by Bespoke Labs' Alex Dimakis, and NVIDIA AI reposted it. Bespoke Labs thanked NVIDIA for early access and collaboration on the work.