{"version":1,"type":"story","url":"https://digestai.news/story/vllm-semantic-router-releases-27b-open-multimodal-decision-model-d3-ul","json":"https://digestai.news/story/vllm-semantic-router-releases-27b-open-multimodal-decision-model-d3-ul.json","markdown":"https://digestai.news/story/vllm-semantic-router-releases-27b-open-multimodal-decision-model-d3-ul.md","slug":"vllm-semantic-router-releases-27b-open-multimodal-decision-model-d3-ul","headline":"vLLM Semantic Router releases 27B open multimodal decision model d3-ultra","summary":"Unlike generative models, d3-ultra does not produce text but assigns probabilities to answers across multiple questions in a single input. It supports text, JSON, images, and videos, with a context window of **131,072 tokens** per question and a **0.46B-parameter vision encoder** included in its **26.09B total parameters**.\n\nThe model excels in text-based decision tasks, scoring **87.2** on the team’s **SR-Text** benchmark (chance-corrected accuracy), surpassing competitors like **GLiDE V1.2 (85.1)** and its sibling **d3-turbo (83.1)**. However, it underperforms in visual grounding (**58.2%** on **SR-Box**), trailing **d3-turbo (60.0%)**. Deployment options include **Hugging Face Transformers**, the **vLLM Semantic Router Engine**, a **free API at decision.vllm-sr.ai**, and a **demo on Hugging Face Space**. The team reports **304 ms latency** for a single photo question on an **AMD Instinct MI325X** GPU, with **2.16 seconds** for **255 questions**—**7.3 ms per additional question**—using **BF16 weights**. No quantized variants are listed yet.","keyPoints":["vLLM Semantic Router released d3-ultra, a 27B open multimodal decision model under Apache-2.0","Supports text, JSON, images, and videos with a 131,072-token context window and 0.46B vision encoder","Scores 87.2 on SR-Text benchmark, but trails d3-turbo in visual grounding (58.2% vs 60.0%)"],"whyItMatters":"d3-ultra enables developers to route, classify, and ground decisions in one API call without text generation, reducing latency for batch processing of up to 255 questions per second. Its open weights and strong text accuracy could accelerate automation in tools requiring probabilistic decision-making.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["vLLM Semantic Router","Marktechpost AI Media Inc."],"models":["d3-ultra","GLiDE V1.2","d1-ultra","d3-turbo","d3","Jev 1.14"],"people":["Asif Razzaq"]},"firstPublishedAt":"2026-10-27T05:50:57Z","updatedAt":"2026-10-27T05:50:57Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"MarkTechPost","title":"vLLM Semantic Router Releases d3-ultra: A 27B Open Multimodal Decision Model","url":"https://marktechpost.com/2026/10/26/vllm-semantic-router-d3-ultra-27b-multimodal-decision-model","publishedAt":"2026-10-27T05:50:57Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":{"title":"vLLM Semantic Router Launches Open Multimodal Decision Models","url":"https://digestai.news/thread/vllm-semantic-router-releases-decision-3-0-multimodal-models-with-5-sizes-under","storyCount":2},"cite":{"text":"Digest AI, \"vLLM Semantic Router releases 27B open multimodal decision model d3-ultra\", 27 October 2026, https://digestai.news/story/vllm-semantic-router-releases-27b-open-multimodal-decision-model-d3-ul","publisher":"Digest AI","title":"vLLM Semantic Router releases 27B open multimodal decision model d3-ultra","datePublished":"2026-10-27T05:50:57Z","url":"https://digestai.news/story/vllm-semantic-router-releases-27b-open-multimodal-decision-model-d3-ul"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}