# wps-decision-1 weights, inputs and deployment

> wps-decision-1 is a multimodal d3-flash fine-tune with 8.394B parameters. See its public weights, score-level limits and deployment details.

Updated 2026-11-05 (Australia/Adelaide)

By Kingsoft Office · Public release: 2026-11-04 · OPEN

A d3-flash fine-tune with typed choice, yes/no and score answers for text, JSON, images and videos. It keeps the Qwen3.5-9B multimodal architecture and decision readout. The bundled Transformers runtime extends score questions to 255 ordered levels; the documented vLLM build keeps the upstream 10-level limit.

## Facts

- Size: 8.394B, including a 0.456B vision encoder
- Licence: Apache-2.0; LICENSE and NOTICE shipped
- Inputs: Text, JSON, images and videos
- Context: 262,144 configured positions; this is not a GPU memory guarantee
- Access: Full BF16 weights and FP32 readout; Transformers or /v1/systemone on the pinned vLLM nightly

## Claims and sources

- **VENDOR-REPORTED**: The card reports a validated deployment on one RTX 5090 with a pinned vLLM nightly and recorded dependencies. Each question gets its own forward pass, batched by the runtime. It publishes no task-accuracy benchmark. [source](<https://huggingface.co/Kingsoft-Office/wps-decision-1>)

## Links

- [Source: Model card and weights](<https://huggingface.co/Kingsoft-Office/wps-decision-1>)
- [Source: Release history](<https://huggingface.co/api/models/Kingsoft-Office/wps-decision-1/commits/main>)

## Related projects

No related projects are listed in the registry.