Unverified50% confidenceFactExact time
DeepSeek already proposed during the V3.1 period that UE8M0 FP8 was designed for next-generation domestic chips; this time it is fully implemented in DeepInfer, meaning chips that do not support this format will directly suffer in precision or performance.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/25/2026
First Seen
Valid until: 1/23/2027
Sources
Related Entities
Related Claims
UnverifiedDeepInfer sets three expectations for hardware: expert weights in FP4, activations and KV in FP8, and scaling factors in the low-precision UE8M0 format; KV Cache per token under 1KB; and the bottleneck shifting to large-scale high-speed interconnect and device-side communication.79% similarUnverifiedAny chip platform that adapts this open-sourced set of components from DeepSeek will be able to run DeepSeek's models.73% similarUnverifiedThere is currently no public information on which hardware DeepInfer supports, whether it is compatible with other models, or how it actually performs compared with mainstream inference frameworks.72% similarUnverifiedDeepSeek's previous models were trained using Nvidia H800 chips before export restrictions were implemented.71% similarUnverifiedDeepSeek的R1模型是在美国禁售前用英伟达H800芯片训练的71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1095895API
curl https://kongchang.com/api/v1/knowledge/claims/1095895MCP
get_claim(id=1095895)