Seed-Coder
Visit ToolSeed-Coder is an open-source code LLM family that includes base, instruct, and reasoning models. It enables LLMs to curate code training data with minimal human effort, enhancing coding capabilities.
At a glance
Trending
Seed-Coder is an open-source code LLM family that includes base, instruct, and reasoning models. It enables LLMs to curate code training data with minimal human effort, enhancing coding capabilities.
Trending
About
Seed-Coder, developed by ByteDance Seed, is a family of lightweight yet powerful open-source code LLMs. It comprises base, instruct, and reasoning models, all of 8B size. A key differentiator is its model-centric approach, predominantly leveraging LLMs for code data filtering and curation, significantly minimizing manual effort in pretraining data construction. Seed-Coder aims to enhance coding capabilities by allowing LLMs to effectively curate their own training data. The project openly shares detailed insights into its model-centric data pipeline, covering GitHub data, commits data, and code-related web data. It achieves state-of-the-art performance among open-source models of comparable size across various coding tasks, including code generation, completion, editing, reasoning, and software engineering.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending