Hi NVIDIA Developer Team,
I’ve been heavily using build.nvidia.com over the last few days for a
personal multi-model benchmarking and agentic workflow project. The 40 RPM
ceiling has become a bottleneck during parallel evaluations — I routinely
launch 10-15 concurrent calls across Kimi K2.6, GLM-5.1, DeepSeek V4 Pro,
Step-3.5-Flash, and Qwen 3.5 397B to compare answer quality on reasoning
and logic prompts. The 429s start hitting after about 4-5 minutes of any
sustained batch run.
Could you bump my account from 40 RPM → 200 RPM? Use case is hobby-grade
research, no commercial scraping, with full attribution to NVIDIA NIM in
any output.Happy to share account details and key ID via private message if you need them.
Thanks a lot!