The vLLM Semantic Router team released Decision 3.0 (d3), an Apache-2.0 family of decision models from 0.6B to 27B built on Qwen3.5 and Qwen3.8-27B. The models output probabilities for choice, yes/no, and score questions without text generation, reading text, images, and video. The 27B ranks #1 on both Jev Decision Index boards, and every size leads its class.
Liquid AI released two open-weight decision models, d1-3B and d1-omni-600M, on Hugging Face. Unlike generative LLMs, they answer in a single forward pass without producing tokens. d1-3B tops the Decision Index 0.2.1 under 10B and answers questions in 16 ms on a Jetson AGX Thor.
Liquid AI released two open-weight decision models, d1-3B and d1-omni-600M, on Hugging Face. Unlike generative LLMs, they answer in a single forward pass without producing tokens. d1-3B tops the Decision Index 0.2.1 under 10B and answers questions in 16 ms on a Jetson AGX Thor.
EmpirioLabs released Aplomb 1, a 5.3B open-weights decision model built on Qwen3.5-4B that reads up to 1M tokens of text, JSON, images, video and audio in a single request. It returns probabilities for every tool and each enum/boolean argument, letting agents act on confident calls and defer uncertain ones to larger models. It ranks #1 on the Hugging Face Typed Decisions benchmark and #1 among 4B models on Decision Index 0.2.1.
EmpirioLabs released Aplomb 1, a 5.3B open-weights decision model built on Qwen3.5-4B that reads up to 1M tokens of text, JSON, images, video and audio in a single request. It returns probabilities for every tool and each enum/boolean argument, letting agents act on confident calls and defer uncertain ones to larger models. It ranks #1 on the Hugging Face Typed Decisions benchmark and #1 among 4B models on Decision Index 0.2.1.
The vLLM Semantic Router project released Decision-2.0-Lux-9B, a decision model fine-tuned from Qwen3.5-9B. Given a state and a set of typed inputs, it answers six questions in a single pass. For practitioners it offers a compact specialized model for routing/decision tasks rather than general-purpose generation.
The vLLM Semantic Router project released Decision-2.0-Lux-9B, a decision model fine-tuned from Qwen3.5-9B. Given a state and a set of typed inputs, it answers six questions in a single pass. For practitioners it offers a compact specialized model for routing/decision tasks rather than general-purpose generation.
Viveki-0.1 is a 20B 'decisions model' that natively handles video, image, and text inputs in real time, with low latency on a single A100 GPU. It was built without any training or fine-tuning, using a claimed mechanistic insight to predict model answers before tokens are generated. It targets developers needing multimodal decision-making where standard LLMs lack video/image processing.
Viveki-0.1 is a 20B 'decisions model' that natively handles video, image, and text inputs in real time, with low latency on a single A100 GPU. It was built without any training or fine-tuning, using a claimed mechanistic insight to predict model answers before tokens are generated. It targets developers needing multimodal decision-making where standard LLMs lack video/image processing.
A developer released MK-SH-V, a 260M-parameter visual decision model built on NeoMME-260M, under open licenses on GitHub and Hugging Face. The model scores candidate decisions from images without generating text and posts 82.6% accuracy on a GUI benchmark versus 73.1% for a 4B zero-shot baseline, with code, weights, and experiments public.
A developer released MK-SH-V, a 260M-parameter visual decision model built on NeoMME-260M, under open licenses on GitHub and Hugging Face. The model scores candidate decisions from images without generating text and posts 82.6% accuracy on a GUI benchmark versus 73.1% for a 4B zero-shot baseline, with code, weights, and experiments public.