# AMD and vLLM Semantic Router release open multimodal model d3-ultra

> Original title: Great work from @AMD and the vLLM Semantic Router team on d3-ultra! Open multimodal decision-making with per-answer probabilities, visual...

- Source: vLLM
- Published: 2026-10-27T03:23:26.000Z
- Bad Influence: https://badinfluence.news/items/ul6al9731qir52qi979nlagc2
- Original: https://x.com/vllm_project/status/2114920881525051889

## Summary

AMD and the vLLM Semantic Router team released d3-ultra, an open multimodal decision model trained and served on AMD Instinct GPUs. The model provides per-answer probabilities, visual grounding, and video tracking, with Apache-2.0 weights available on Hugging Face. According to vLLM's background post, d3-ultra is a 27B model that ranks #2 of 230 on the Jev Decision Index text board, 3.03 points above d3.

## Post

Great work from @AMD and the vLLM Semantic Router team on d3-ultra! Open multimodal decision-making with per-answer probabilities, visual grounding, and video tracking—trained and served on AMD Instinct GPUs.
Apache-2.0 weights on @huggingface.

## Quoted post @XunzhuoLiu

> Meet d3-ultra: our newest open 27B multimodal decision model.
> 
> #2 of 230 on the Jev Decision Index text board, +3.03 over d3.
> Text, images, video: it boxes what it finds and tracks it through time.
> 
> Open weights (Apache-2.0): https://huggingface.co/vllm-sr/d3-ultra
> 
> #Jev #Decision

https://x.com/XunzhuoLiu/status/2114915380926017591