# Cline nearly doubles Nemotron 3.5 Super VL throughput on one GB300 GPU

> Original title: We nearly doubled throughput for Nemotron 3.5 Super VL on a single GB300 GPU, from 6,058 → 11,840 tok/s, with five configuration changes ...

- Source: Cline
- Published: 2026-10-29T19:13:55.000Z
- Bad Influence: https://badinfluence.news/items/ujcymk9o0ysq4xu2hd59i7q86
- Original: https://x.com/cline/status/2115884854139380134

## Summary

Cline says five vLLM configuration changes raised Nemotron 3.5 Super VL throughput on a single GB300 GPU from 6,058 to 11,840 output tokens per second. The changes let the model draft tokens ahead, keep the GPU from sitting idle, and store the model's memory at half precision. The chart shows the gains step by step: MTP decoding reached about 7,000, async with FlashInfer 7,800, MTP 3 changed to MTP 1 9,500, and FP16 state 11,840 at 256 concurrent requests.

## Why it matters

The post lists five vLLM configuration changes with throughput measured after each step, showing how each contributes to the gain on one GPU.

## Post

We nearly doubled throughput for Nemotron 3.5 Super VL on a single GB300 GPU, from 6,058 → 11,840 tok/s, with five configuration changes in vLLM.

These let the model draft tokens ahead, keep the GPU from ever sitting idle, and store the model’s memory at half precision.

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHV0g3SJbsAAD7FQ.jpg&mode=thumb&exp=1793491200&sig=58c4e858445999e0)