Unverified50% confidenceFactExact time
A 2B model requires approximately 4–10 TFLOPs of computation per inference, taking 10–50 milliseconds on an A100 GPU, while a 70B model requires 140–350 TFLOPs and can take up to several seconds.
1
Sources
50%
Confidence
Long-term
Relevance
9/9/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAt FP16 half-precision computation, the RTX 4090 has a theoretical performance of approximately 330 TFLOPS, while the M2 Ultra's GPU delivers approximately 27.2 TFLOPS.73% similarUnverifiedMainstream GPU inference solutions (such as deployments based on NVIDIA H100) typically achieve only a few dozen to one or two hundred tokens per second on 27B-scale models.72% similarUnverifiedBecause 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.69% similarUnverifiedIn real-world testing on an RTX 3060 (6GB VRAM) with 32GB RAM running the pre-quantized Krea2 Turbo, 1024×1024 resolution with 8-step Euler sampling achieved a speed of 1.78 s/it and a total elapsed time of 17.64 seconds.68% similarUnverified测试者用双4090（48GB显存）工作站运行20.7B稠密模型，每秒生成27.6个Token，仅比矮星略快，但参数量小10倍以上68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/882138API
curl https://kongchang.com/api/v1/knowledge/claims/882138MCP
get_claim(id=882138)