← All models

Seed1.6-Embedding

Published Publisher ByteDancedoubao-embedding-vision-250615

Positioning and Release Date

Seed1.6-Embedding is a multimodal embedding model based on Seed1.6-Flash, designed to encode content as vectors rather than generate conversational responses. ByteDance's official announcement was dated 2025-06-28 and explicitly listed the Volcano Ark model ID doubao-embedding-vision-250615. (seed.bytedance.com)

Capabilities and Uses

  • Supports text, image, video, and mixed inputs, and is suitable for cross-modal retrieval, text-to-image search, image-to-image search, mixed image-text search, and semantic matching.
  • Uses a dual-tower architecture; dense vectors support 2048 or 1024 dimensions. The official model list specifies a 128K context window.
  • Volcano Ark documentation states that this version also supports sparse vectors, but only for text-only input. (seed.bytedance.com)

Limitations and Service Status

This version does not support multi-vector output, custom video frame extraction, or the instructions field; sparse vectors do not support image or video input. Volcano Ark has listed it as a version scheduled for deprecation: new purchases will stop on 2026-09-24, and migration to doubao-embedding-vision-251215 is recommended; the announcement states that switching is not mandatory for this model. Before migration, performance should be evaluated and existing vector knowledge bases rebuilt. (console.volcengine.com)

Pricing

Date checked: 2026-11-18.

  • Mainland China, Volcano Ark, Chinese yuan (CNY): Official pricing is listed for the doubao-embedding-vision model series at 0.70 yuan/million tokens for text input and 1.80 yuan/million tokens for image input; billing is based on input tokens, with no separate vector output fee listed. Image tokens are calculated according to official rules. Actual availability for purchase and fees are subject to the platform. (console.volcengine.com)
  • International site: BytePlus lists pricing for the skylark-embedding-vision series at 0.125 USD/million tokens for text input and 0.325 USD/million tokens for image input, billed based on input, with no separate vector output fee listed; its documentation also lists a 250615 version. However, it has not been independently verified that this international-site model is exactly the same model as the domestic API ID in this entry, so this price cannot be treated as a direct USD quote for that domestic ID. (docs.byteplus.com)

Official Sources