Unverified60% confidenceEventExact time
According to social media leaks, DeepSeek is training a model with 2 trillion (2T) parameters and plans to ultimately build an ultra-large-scale model with 8 trillion (8T) parameters.
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
9/21/2026
First Seen
Valid until: 12/20/2026
Sources
Related Entities
Related Claims
UnverifiedIn November 2023, DeepSeek open-sourced the DeepSeek LLM and Coder models, with the largest model having 67 billion parameters and trained on 2 trillion tokens.80% similarUnverifiedDeepSeek Pro's 1.6T total parameters with 49B activated parameters result in inference costs comparable to a roughly 49B dense model, while expanding knowledge capacity by approximately 32 times.69% similarUnverifiedIn the training camp, Xiao Bin demonstrated an enterprise-level project that simultaneously connects to multiple large model vendors: the main reasoning model uses the DeepSeek series, the summarization model uses a lightweight version, and a Zhipu model is configured as a backup reasoning channel.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/938983API
curl https://kongchang.com/api/v1/knowledge/claims/938983MCP
get_claim(id=938983)