Unverified50% confidenceFactExact time
DeepSeek V4 Flash 0423 has an input discount of 90% and an output discount of 95% on UplinkSwitch, the largest discount tier in the entire dataset.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/17/2026
First Seen
Valid until: 10/31/2026
Sources
Related Entities
Related Claims
Partially VerifiedOn a dual-Spark system, DeepSeek V4 Flash (FP8 quantization) peaks at approximately 53 tokens per second.71% similarUnverifiedThe DeepSeek Flash model can output more than 100 tokens per second.68% similarUnverifiedFlashMLA's prefill throughput for the sparse attention operator on the Ascend 950 reaches 410 TFLOPS (95% of the hardware's theoretical peak), and decode reaches 360 TFLOPS (83%), both figures as published by DeepSeek itself with no third-party measurements.68% similarUnverified盘古2.0 Flash在LiveCodeBench算法推理测试中得分85.1，仅次于DeepSeek V4 Flash的91.6，超过千问3.6 27B的83.967% similarVerifiedDeepSeek-V4.1-Flash's KV cache is compressed to 890 bytes per token, only one-quarter that of V4 Flash.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1057008API
curl https://kongchang.com/api/v1/knowledge/claims/1057008MCP
get_claim(id=1057008)