Now Reading: ByteDance Trains AI Model With Up To 10 Trillion Parameters
-
01
ByteDance Trains AI Model With Up To 10 Trillion Parameters
ByteDance is training an artificial intelligence model with as many as 10 trillion parameters, the Financial Times reports, potentially putting it near Anthropic’s estimated Mythos 5 scale. Reuters could not independently verify…
The Story in Brief
ByteDance is training an artificial intelligence model with as many as 10 trillion parameters, the Financial Times reports, potentially putting it near Anthropic’s estimated Mythos 5 scale. Reuters could not independently verify the report, and ByteDance did not immediately comment.
The model would be more than three times the size of Moonshot AI’s 2.8 trillion-parameter Kimi K3. Chinese models from Meituan and DeepSeek previously had about 1.6 trillion parameters. The ByteDance system is in pre-training, which typically lasts three to six months before fine-tuning and release. Parameter counts offer a rough measure of model scale, not capability. Anthropic and OpenAI do not disclose counts for several leading systems, making direct comparisons difficult.
The Indian Opinion
The loudest narrative will cast this as proof that China has overtaken US AI, while another may dismiss it as a race for vanity numbers. Both go beyond the evidence. This is a reported project, not a released product, and parameter size does not establish performance, cost or reliability. The useful test is whether ByteDance publishes results, access details and operating costs after pre-training, rather than just a headline figure of 10 trillion.
Sources (2): timesnownews.com, livemint.com
This story was synthesised by AI from the 2 sources linked above.
