Social media giant ByteDance on Wednesday released a new artificial intelligence model that looks to challenge those from DeepSeek and Alibaba Cloud in the global market for open-source AI systems.
TikTok and Douyin owner ByteDance’s Seed department, which oversees the tech unicorn’s AI research and large language model (LLM) development, has uploaded three variants of its Seed-OSS-36B open-source AI model to Hugging Face, the world’s largest open-source AI developer platform.
Seed-OSS-36B was designed with both general-purpose and reasoning capabilities, with support for long-context window processing and various developer-friendly features, ByteDance said in a statement.
Do you have questions about the biggest topics and trends from around the world? Get the answers with SCMP Knowledge, our new platform of curated content with explainers, FAQs, analyses and infographics brought to you by our award-winning team.
Despite a modest size of 36 billion parameters – the internal variables used by the system during training to improve its performance – Seed-OSS-36B offered competitive capabilities, the company said. Open-source AI models released by other Chinese companies have parameters dozens of times larger.
Citing recent benchmark tests, ByteDance said its new AI model either matched or surpassed the capabilities of similar-sized offerings from competitors including Alibaba Cloud, Google and OpenAI, as well as the firm’s own Seed 1.6 model. Alibaba Cloud is the AI and cloud computing arm of Alibaba Group Holding, owner of the South China Morning Post.
ByteDance’s Seed department oversees the social media giant’s artificial intelligence research and large language model development. Photo: Shutterstock alt=ByteDance’s Seed department oversees the social media giant’s artificial intelligence research and large language model development. Photo: Shutterstock>
ByteDance’s release of Seed-OSS-36B AI reflects how Chinese AI companies are continuing to narrow the gap with their US peers through the open-source approach, which makes the source code of AI models available for third-party developers to use, modify and distribute.
The Beijing-based company made its foray into AI models in May last year with the release of its proprietary Doubao LLM family, which shares the same name as the Doubao chatbot that it introduced in 2023. The firm’s flagship Seed 1.6 model remains a closed-source system.
The launch of an open-source model comes more than a year after ByteDance CEO Liang Rubo criticised employees for reacting too slowly to fast-developing trends in the market, especially in AI.
As of July, China was home to the world’s top open-source AI models, according to a report by LMArena, a benchmarking platform created by researchers from the University of California, Berkeley.
Moonshot AI’s Kimi K2, MiniMax M1, Alibaba Cloud’s Qwen 3 and a variant of DeepSeek’s R1 were ranked as the world’s top open-source AI models, beating out offerings like Google’s Gemma 3-72B and Meta Platforms’ Llama 4-Maverick, the LMArena report said.
This article originally appeared in the South China Morning Post (SCMP), the most authoritative voice reporting on China and Asia for more than a century. For more SCMP stories, please explore the SCMP app or visit the SCMP’s Facebook and Twitter pages. Copyright © 2025 South China Morning Post Publishers Ltd. All rights reserved.
Copyright (c) 2025. South China Morning Post Publishers Ltd. All rights reserved.