Unverified50% confidenceFactExact time
Flash-series models typically adopt a parameter scale of under 70B.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/8/2026
First Seen
Valid until: 12/7/2026
Sources
Related Entities
Related Claims
UnverifiedBecause Flash models have fast inference speeds (50–200ms per request vs. 500–2000ms for flagship models), they can support higher QPS under the same TPM limits.71% similarUnverifiedThe approximately 20:1 ratio between Qwen3 Next Flash's 125 billion total parameters and 6 billion activated parameters represents a fairly aggressive sparsity design.68% similarUnverifiedThe DeepSeek Flash model can output more than 100 tokens per second.66% similarUnverifiedPangu 2.0 Flash happens to fill the current gap in the open-source community between 27B and 200B, as there was previously no open-source MoE model of approximately 90B.66% similarUnverifiedFull fine-tuning of a 70B-scale model requires at least 8 A100 80G GPUs.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/879171API
curl https://kongchang.com/api/v1/knowledge/claims/879171MCP
get_claim(id=879171)