Seed2.0 Code
Positioning and capabilities
Seed2.0 Code is a programming-focused preview model in the Seed2.0 family. ByteDance describes this version as optimized for enterprise coding, including front-end development and programming in multiple languages, with agent and visual-understanding capabilities. It is intended for integration into AI coding tools. (docs.volcengine.com)
Release and access
The official Seed2.0 series announcement is dated February 14, 2026 and identifies Code as a dedicated model in the family. The 260215 suffix is part of the API snapshot identifier; it does not independently establish that snapshot’s release date. The Volcengine model release announcement lists the exact API ID under February 2026, without establishing a specific day for that snapshot. (seed.bytedance.com)
The official series announcement says Code was available through Doubao and TRAE, with the series API on Volcengine. The official model list marks this exact preview as scheduled for retirement; verify current availability before building a dependency on it. The weight-release status of this specific version is unverified: a hosted API alone establishes neither open-weight nor closed-source status. (seed.bytedance.com)
Pricing
Mainland China — Volcengine Ark: The official pricing page, checked November 17, 2026, lists the model family as doubao-seed-2.0-code, rather than the full snapshot ID. The following are therefore family-level reference prices; confirm applicability to this specific preview and the final order before use. Rates are in CNY per million tokens, with tiers determined by input length: (console.volcengine.com)
| Input length | Standard input | Cached input | Standard output |
|---|---|---|---|
| Up to 32K tokens | 3.2 | 0.64 | 16.0 |
| Over 32K to 128K | 4.8 | 0.96 | 24.0 |
| Over 128K to 256K | 9.6 | 1.92 | 48.0 |
For batch inference, the same page lists input/output rates of CNY 1.6/8.0, 2.4/12.0 and 4.8/24.0 per million tokens for the respective tiers; cached-input rates are 0.64, 0.96 and 1.92. Batch and standard online inference are different billing modes. (console.volcengine.com)
An official international USD input/output price for this exact snapshot could not be verified on the BytePlus ModelArk pricing page, checked November 17, 2026. No currency conversion or third-party resale price has been substituted. (docs.byteplus.com)
Limitations
This is a preview version and is marked for retirement. Generated code and agent actions should be reviewed and tested before deployment; the series announcement also acknowledges remaining challenges in end-to-end code generation and contextual learning. (docs.volcengine.com)