ByteDance: Doubao-Seed-2.0-lite API
Doubao-Seed-2.0-Lite (v260215) is the high-efficiency, cost-optimized branch of the Doubao 2.0 foundation model family. Built on a refined Mixture-of-Experts (MoE) architecture, it achieves a 3x speedup in time-to-first-token (TTFT) compared to the Pro version, while retaining nearly 90% of its reasoning capabilities. In 2026, it is specifically optimized for "Lightweight Thinking," supporting real-time deep reasoning and multimodal understanding at a fraction of the cost. It is an ideal engine for large-scale customer service agents, real-time content moderation, and high-frequency automated tools, delivering a premium intelligent experience for millions of concurrent users.
- Context window: 256,000 tokens
- Max output: 128,000 tokens
- Input: text, image, video, file
- Output: text
- Reasoning: Supported
- Tool calling: Supported
- File input: Supported
- Released: 2026-02-13
- Knowledge cutoff: 2026-01-31
Frequently Asked Questions
What is the context window of Doubao-Seed-2.0-lite?
Doubao-Seed-2.0-lite supports a context window of up to 256,000 tokens.
Does Doubao-Seed-2.0-lite support function calling?
Yes. Doubao-Seed-2.0-lite supports tool / function calling.
Does Doubao-Seed-2.0-lite support reasoning?
Yes. Doubao-Seed-2.0-lite is a reasoning-capable model.
What is the knowledge cutoff of Doubao-Seed-2.0-lite?
The knowledge cutoff of Doubao-Seed-2.0-lite is 2026-01-31.