Japanese Giant AI Model Accused of Copying DeepSeek

A high-profile launch in Japan's tech sector has ignited a public debate about technological transparency. Rakuten Group recently unveiled what it describes as the "largest and most powerful" large language model developed in-house. This 70-billion-parameter model is a major outcome of a project supported by Japan's Ministry of Economy, Trade and Industry (METI) under its GENIAC initiative. Shortly after its release, however, the open-source community began questioning if it was merely a repackaged version of an existing model.
Developers discovered that the model's core architecture and configuration file pointed unmistakably to a model created by a Chinese team. Evidence indicates Rakuten's model not only retained the original identifier "DeepseekV3ForCausalLM" but was in fact a fine-tuned adaptation using Japanese data, not a ground-up development.
The controversy revolves around what many see as a "gray area" in Rakuten's communications and license management:
Inadequate Disclosure: The official press release only mentioned "integrating the essence of the open-source community," failing to acknowledge the foundational model's origin.
License Management Issues: The community highlighted that Rakuten's initial release may have omitted the required MIT license files. While a NOTICE file was added later to comply with legal terms, this retroactive fix was criticized as lacking openness and genuine intent.
To date, Rakuten Group has not directly addressed the removal of license files or the striking architectural similarities.
Related article
SpaceX IPO Filing Highlights Satellite Internet and AI Expansion Ambitions
In its S-1 registration statement filed ahead of a planned IPO, SpaceX recently unveiled a number of impressive business metrics that highlight its strong footprint in aerospace communications and artificial intelligence:Starlink subscribers surpass
Alibaba Tuhao M890 Debuts with Triple Performance, Ushering in Full-Stack Agent Era for Chip-Cloud-Model-Inference
On May 20, 2026, at the Alibaba Cloud Summit, Alibaba Cloud announced the completion of a full-stack technology system upgrade designed for the Agentic era. The transformation reshaped the entire pipeline—from underlying chips and cloud platform to m
Pentium 4 Revival: 20-Year-Old CPU Runs Meta Llama 3 Large Model
Recently, the YouTube tech channel Fully Buffered carried out an impressive and hardcore experiment: successfully running Meta's latest Llama 3.2 3B large model on the Pentium 4 641 processor, a chip released in 2006.This test forced modern artificia
Related Special Topic Recommendations
Comments (0)
0/500

A high-profile launch in Japan's tech sector has ignited a public debate about technological transparency. Rakuten Group recently unveiled what it describes as the "largest and most powerful" large language model developed in-house. This 70-billion-parameter model is a major outcome of a project supported by Japan's Ministry of Economy, Trade and Industry (METI) under its GENIAC initiative. Shortly after its release, however, the open-source community began questioning if it was merely a repackaged version of an existing model.
Developers discovered that the model's core architecture and configuration file pointed unmistakably to a model created by a Chinese team. Evidence indicates Rakuten's model not only retained the original identifier "DeepseekV3ForCausalLM" but was in fact a fine-tuned adaptation using Japanese data, not a ground-up development.
The controversy revolves around what many see as a "gray area" in Rakuten's communications and license management:
Inadequate Disclosure: The official press release only mentioned "integrating the essence of the open-source community," failing to acknowledge the foundational model's origin.
License Management Issues: The community highlighted that Rakuten's initial release may have omitted the required MIT license files. While a NOTICE file was added later to comply with legal terms, this retroactive fix was criticized as lacking openness and genuine intent.
To date, Rakuten Group has not directly addressed the removal of license files or the striking architectural similarities.
SpaceX IPO Filing Highlights Satellite Internet and AI Expansion Ambitions
In its S-1 registration statement filed ahead of a planned IPO, SpaceX recently unveiled a number of impressive business metrics that highlight its strong footprint in aerospace communications and artificial intelligence:Starlink subscribers surpass
Alibaba Tuhao M890 Debuts with Triple Performance, Ushering in Full-Stack Agent Era for Chip-Cloud-Model-Inference
On May 20, 2026, at the Alibaba Cloud Summit, Alibaba Cloud announced the completion of a full-stack technology system upgrade designed for the Agentic era. The transformation reshaped the entire pipeline—from underlying chips and cloud platform to m
Pentium 4 Revival: 20-Year-Old CPU Runs Meta Llama 3 Large Model
Recently, the YouTube tech channel Fully Buffered carried out an impressive and hardcore experiment: successfully running Meta's latest Llama 3.2 3B large model on the Pentium 4 641 processor, a chip released in 2006.This test forced modern artificia





Home






