Unverified50% confidenceBenchmarkExact time
开发者公布在RTX 3050 Laptop（4GB）上微调Llama-3.1-8B的实测结果为训练速度119.6 tok/s，峰值显存3.32 GB
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/12/2026
First Seen
Valid until: 11/10/2026
Sources
Soup CLI：4GB显卡微调80亿参数大模型的开源工具详解
Related Claims
UnverifiedIn 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.80% similarUnverifiedOn the RTX 5090, the Qwen3.6 35B-A3B achieves an inference speed of 220 tokens/s with approximately 23 GB of VRAM usage.73% similarUnverifiedOn the RTX 5090, the Qwen3.6 27B model's prefill speed increases from 174 to 253 tokens/s, with a VRAM usage of approximately 18 GB.72% similarUnverifiedThe RTX 4060 can achieve generation speeds 3-5 times that of the GTX 1060.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/732848API
curl https://kongchang.com/api/v1/knowledge/claims/732848MCP
get_claim(id=732848)