Unverified50% confidenceFactExact time
Meta的Llama 3.1提供8B/70B/405B尺寸，Google的Gemma系列提供2B/7B/27B尺寸
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/24/2026
First Seen
Valid until: 11/22/2026
Sources
Related Entities
Related Claims
UnverifiedLlama 3 8B has 32 layers, 8 KV heads (GQA architecture with 32 query heads), a head dimension of 128, and a KV cache of approximately 1.07 GB for a single sequence under FP16 with an 8192 context length.67% similarUnverifiedTencent Hunyuan released and open-sourced the official version of the Hunyuan3 large model, which adopts a MoE architecture with a total of 295B parameters, 21B activated parameters, and supports a 256K context window.66% similarUnverifiedFull fine-tuning of Llama-3 (8B) requires approximately 80 GB or more of VRAM, while Llama-3 (70B) requires more than 400 GB.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/793781API
curl https://kongchang.com/api/v1/knowledge/claims/793781MCP
get_claim(id=793781)