Positioning
Seed-Evolving is a continuously updated deep-thinking model designed for agent and coding tasks. Its unified Model ID automatically serves the latest version rather than a fixed release. (console.volcengine.com)
Capabilities and uses
It supports complex task orchestration, long-horizon planning, code generation, tool calling, text generation, multimodal understanding, GUI tasks and structured output. The August 27 update strengthened general agents, coding, deep search, tool use and visual understanding. Suitable uses include software development, research and workflow automation. (console.volcengine.com)
Limits and considerations
- The current Volcengine model list states a 1024K-token context window, a 1024K-token maximum input and a 256K-token maximum answer; the default answer limit is 4K tokens. These are current specifications, not independently verified historical limits for the August release. (console.volcengine.com)
- Automatic updates can change model behavior. The provider recommends returning encrypted reasoning content as instructed and keeping configuration consistent within a session. Outputs and tool results should still be checked for errors. (console.volcengine.com)
- BytePlus documentation identifies this model ID as not currently available on BytePlus. (ai.byteplus.com)
Release dates
The model first launched on June 23, 2026. August 27, 2026 is an update date, not its initial release date. The official version history lists November 5, 2026 as the latest update at the time checked. (console.volcengine.com)
Pricing
Volcengine China, standard online inference, pay-as-you-go, for input length 0–1024K tokens: non-audio input CNY 6.00 per million tokens; output CNY 30.00 per million tokens; non-audio cache-hit input CNY 1.20 per million tokens; cache storage CNY 0.017 per million tokens per hour. These are current published rates, not verified historical August rates. Checked November 16, 2026. Official pricing. (docs.volcengine.com)