Unverified50% confidenceFactExact time
Meta's LLaMA 3 series has a maximum scale of 405B, and DeepSeek-V3 is approximately 671B (mixture-of-experts architecture).
1
Sources
50%
Confidence
Long-term
Relevance
7/24/2026
First Seen
Sources
小米静默发布MiMo-V2.5-DFlash：300B模型推理速度或将翻倍
redditr/LocalLLaMA7/12/2026
Related Claims
UnverifiedMeta's Llama 3.1 comes in multiple parameter versions ranging from 8B to 405B, covering full-scenario needs from edge deployment to data centers.71% similarUnverifiedHigh-quality open-source models such as Llama 3.1 405B and DeepSeek V3 have already come very close to commercial flagship models on many general tasks.71% similarUnverifiedMeta's Llama 3.2 series introduced lightweight versions of 1B and 3B parameters.71% similarUnverifiedMeta发布的Llama 3.1已支持405B参数规模，并以宽松的商业授权开放64% similarUnverifiedDeepSeek V4 Pro scored 80.6 on SWE-Verify and 83.4 on BrowsCamp.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/607363API
curl https://kongchang.com/api/v1/knowledge/claims/607363MCP
get_claim(id=607363)