Unverified50% confidenceFactExact time
DeepSeek's open-source Mega MoE kernel previously lacked a backpropagation implementation and could not be used directly for training.
1
Sources
50%
Confidence
Long-term
Relevance
10/29/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPrime Intellect released Prime Mega MoE, adding backpropagation to DeepSeek's fused Mega MoE kernel, making it trainable.74% similarUnverifiedDeepMind used a large-scale distributed training cluster for StarCraft II because a single machine could not complete training within an acceptable timeframe.69% similarUnverifiedWhat has been open-sourced this time are operators and libraries, not models, which does not mean Ascend can directly run DeepSeek services.62% similarUnverifiedSome of DeepSeek's models restrict using their outputs to train competing models.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1112886API
curl https://kongchang.com/api/v1/knowledge/claims/1112886MCP
get_claim(id=1112886)