vLLM Semantic Router Releases d3-ultra: A 27B Open Multimodal Decision Model
vLLM Semantic Router launches d3-ultra, a 27B open-weight multimodal decision model processing text, images, and video, scoring 87.2 on SR-Text.
Stock photo for illustration only, not from the actual event
- vLLM Semantic Router launched d3-ultra, a 27B open-weight model
- Processes text, images, and video simultaneously in a single pass
- Returns probabilities for every possible answer generated
- Achieved a score of 87.2 on the team's SR-Text test
The artificial intelligence community has a new contender with the release of d3-ultra by vLLM Semantic Router. This newly introduced open-weight multimodal decision model packs 27 billion parameters, engineered to elevate the standard for processing heterogeneous data types efficiently and accurately.
A standout characteristic of the d3-ultra model is its capability to ingest text, images, and video simultaneously in a single reading pass. This unified approach eliminates the latency and pipeline complexity typically associated with legacy architectures that require separate processing stages for different media modalities.
When it comes to output generation, d3-ultra delivers probabilities for every potential answer rather than a single deterministic guess, empowering downstream systems to make more nuanced and informed decisions. Furthermore, the model secured an impressive score of 87.2 on the development team's internal SR-Text test.
Stock photo for illustration only, not from the actual event
The introduction of d3-ultra highlights the rapid momentum within the open-source AI ecosystem to bridge the gap between text-based reasoning and rich perceptual inputs. Models capable of native multimodal processing in a single pass are critical for advanced real-time applications such as autonomous navigation, interactive video analysis, and sophisticated AI agents that require instantaneous multi-sensory comprehension.
Source: MarkTechPost
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment