The Beijing-based company claims the model surpasses OpenAI’s o1 on AIME, a benchmark evaluating AI performance on complex tasks.
This release follows Chinese AI startup DeepSeek’s launch of DeepSeek-R1 earlier in the week. DeepSeek-R1, an open-source model, reportedly rivals OpenAI’s o1 in multiple performance benchmarks.
DeepSeek has gained attention for developing competitive AI models with a lower budget and offering significantly reduced user fees.
Doubao-1.5-pro is aggressively priced through ByteDance’s Volcano Engine cloud platform. The 32k version costs 2 yuan (approximately US$0.27) per million tokens, while the more powerful 256k version is priced at 9 yuan (approximately US$1.36).
For comparison, DeepSeek charges 16 yuan (US$2.98) per million tokens for DeepSeek-R1, and OpenAI’s o1 is priced at 438 yuan (US$64.94).
Other Chinese companies, including Moonshot AI, Minimax, and iFlyTek, have also introduced reasoning models in recent weeks, aiming to compete with global players.
OpenAI, which started the AI reasoning race with ChatGPT in 2022 and its “Strawberry” series in 2023, is preparing to launch a new model, o3 mini, in the coming weeks, according to CEO Sam Altman.
These developments indicate growing competition in AI reasoning capabilities and pricing. Chinese companies’ competitive pricing strategies could disrupt market dynamics, particularly for enterprises looking to adopt advanced AI solutions at scale while managing costs.
