decision2
6 TagsUpdated Apache-2.0by vLLM Semantic Router
Decision 2.0 by the vLLM Semantic Router contributors: a fully fine-tuned Qwen3 or Qwen3.5 backbone plus an endpoint head that scores every option at its own last token against the question, in one forward pass per question. Five sizes, 0.6B to 9B.
decisionlong-contextfine-tuned0.60b0.75b1.88b4.21b7.94b
Tags
| Name |
|---|
| decision2latestSame as decision2:nox.9.7 GB · 16384 ctx · English |
| decision2:eosQwen3.5-0.8B: 0.460 on typed decisions.2.0 GB · 16384 ctx · English |
| decision2:kaiQwen3-0.6B-Base, the fastest: 31 ms for five questions on an RTX 5090; 8k-token rows.1.5 GB · 8192 ctx · English |
| decision2:luxQwen3.5-9B, the largest: 0.707 on typed decisions.18.0 GB · 16384 ctx · English |
| decision2:noxQwen3.5-4B-Base: 0.680 on typed decisions; 251 ms for five questions on an RTX 5090.9.7 GB · 16384 ctx · English |
| decision2:solQwen3.5-2B: 0.554 on typed decisions.4.8 GB · 16384 ctx · English |
Tags without a suffix load fp16 on a CUDA GPU and fp32 on CPU; add -fp16 or -fp32 to pin one precision.