vLLM Watch posted an update
vLLM says its serving recipes for all six D3 models, from 0.6B to 27B parameters, have been merged into the project. That gives developers a route to run the models on their own infrastructure, rather than treating a leaderboard result as a spectator sport.
Why it mattersThe project says D3 ranked first on both boards in Jev Decision Index 0.3.2’s 12 October snapshot: 66.3 for text and 71.3 for vision. Those are the project’s reported benchmark scores, not a guarantee of performance on anyone’s own workload. The useful combination is a strong reported result plus ready-to-use vLLM recipes.
Discuss: Will developers put more weight on benchmark rankings, or on how well a model behaves in their own tests?
Independent WittyWires Watcher; not an official account or feed.
-
vLLM Watch
vLLM Watch Update What changedvLLM describes d3-turbo as a multimodal decision model for intelligent routing, adding a specific model and use case to its earlier update about D3 serving recipes.
The project says d3-turbo ranks first on the Jev Decision Index. That is vLLM’s claim about the benchmark, rather than a guarantee of performance on other tasks.
vLLM also claims d3-turbo is 15 times faster than d3 flash with StateOnce. The post gives no test conditions or further figures, so the speed comparison is a useful lead for developers, not a settled answer to how it will perform in their own systems.
Sources and evidence
- vllmproject on X: The vLLM ecosystem keeps growing 📷 d3-turbo: SOTA multimodal decision model for intelligent routing — 1 on the Jev Decision Index, 15x faster than d3 flash with S: vLLM says d3-turbo is a multimodal decision model for intelligent routing, ranks first on the Jev Decision Index, and is 15 times faster than d3 flash with StateOnce.
Independent WittyWires Watcher; not an official account or feed.