Unverified50% confidenceFactExact time
Llama, Qwen, and DeepSeek all offer versions at various parameter scales ranging from 1.5B to over 70B
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/12/2026
First Seen
Valid until: 12/11/2026
Sources
Related Entities
Related Claims
UnverifiedQwen3.8-Max-0902 uses a Mixture-of-Experts (MoE) architecture with 2.4T parameters and supports a context length of 128K.74% similarUnverifiedQwen3.8-Flash-Next has a total parameter count of 125B with approximately 6B activated parameters, and belongs to the sparse MoE architecture.73% similarUnverifiedEarly LLaMA-2 only supported a 4K token context, whereas new-generation models such as DeepSeek-V3 and Qwen2.5 have extended the context to 128K or even longer.73% similarUnverifiedQwen3.5-122B-A10B is a MoE architecture model with approximately 122B total parameters and approximately 10B activated parameters.72% similarUnverifiedWhen running the Qwen3 conversational model, 8GB of VRAM supports the 1.7B/6B parameter versions, while 12GB/16GB/24GB of VRAM supports larger parameter versions.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/912849API
curl https://kongchang.com/api/v1/knowledge/claims/912849MCP
get_claim(id=912849)