Multimodal / Vision

SenseNova 6.8 Code2Video

SenseNova 6.8 Code2Video by SenseTime (SenseNova) has no tracked price. It has no overseas route listed on this record. Last checked 2026-11-02.

SenseTime (SenseNova) · apache-2.0 · 27.4B

—

unpriced · Last updated 2026-11-02 · Pricing source

Code-driven video model that writes executable code to build and render short videos. It does not generate video pixels. Fine-tuned from Qwen3.8-27B; architecture Qwen3_5ForConditionalGeneration and language_model_only is false. Hugging Face pipeline is image-text-to-text, so this row is multimodal. Card badge: 27B parameters and a 512K context. HF safetensors count is 27.357B. Card license is Apache-2.0. The 512K figure is 524288 tokens in the published LightLLM, vLLM, and SGLang commands. config.json text_config.max_position_embeddings is 262144; the vLLM and SGLang examples set 524288 with the long-context override flags. release_date is the Hugging Face repo creation date. The OpenSenseNova GitHub repo was created the same day: https://github.com/OpenSenseNova/SenseNova-6.8-Code2Video. No USD price.