Unverified85% confidenceFactTime unknown
LLaMA系列最大的版本达到70B（700亿参数）
1
Sources
85%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedLLaMA-2 70B 总参数约 700 亿均匀分布于 80 层，每层约 8.75 亿参数，FP16 精度下单层约占 1.75GB 显存72% similarPartially VerifiedTaking the 65B-parameter LLaMA model as an example, traditional full fine-tuning requires more than 780GB of VRAM, whereas QLoRA requires only approximately 48GB.63% similarUnverifiedRunning an open-source model at the 70-billion-parameter scale requires at least approximately 140GB of VRAM in FP16 precision, i.e., 8 H100s in parallel.62% similarUnverifiedThe weight files of LLaMA-70B can reach 140 GB, and a single download consumes enormous bandwidth.61% similarUnverifiedLLaMA 后续版本将词表从 3.2 万扩展至 12.8 万61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/22427API
curl https://kongchang.com/api/v1/knowledge/claims/22427MCP
get_claim(id=22427)