Unverified50% confidenceReleaseExact time
Prime Intellect released Prime Mega MoE, adding backpropagation to DeepSeek's fused Mega MoE kernel, making it trainable.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/29/2026
First Seen
Valid until: 1/27/2027
Sources
Related Entities
Related Claims
UnverifiedDeepSeek's open-source Mega MoE kernel previously lacked a backpropagation implementation and could not be used directly for training.74% similarUnverifiedThe UP host recommends using DeepSeek 4.1 flash or GLM 5.3 flash for general tasks, and Kimi K3 for long-chain animated short drama tasks when credits are sufficient.61% similarUnverifiedMoE models such as DeepSeek and Mixtral adopt the "large total, small activation" naming convention.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1112883API
curl https://kongchang.com/api/v1/knowledge/claims/1112883MCP
get_claim(id=1112883)