Unverified50% confidenceFactExact time
Agnes-3.0-Flash's global attention layers use 6:1 GQA with 24 query heads and 4 KV heads, a head dim of 256, RMS-norm applied to q and k, and sigmoid-gated output
1
Sources
50%
Confidence
Long-term
Relevance
9/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAgnes-3.0-Flash's delta-rule layers are configured with 16 key heads, 48 value heads, head dim 128, a preceding causal convolution with kernel size 4, paired with gated RMS-norm, and recurrent states stored in fp32 precision79% similarUnverifiedAgnes-3.0-Flash has a total of 72 layers, consisting of 54 delta-rule recurrent layers and 18 global attention layers arranged in a 3:1 alternating pattern, with a hidden dimension of 5120 and a vocabulary size of 248,32076% similarUnverifiedThe Ling-3.0-flash model has a total parameter count of 124B, with approximately 5.1B parameters activated per token, employs a KDA + MLA hybrid attention mechanism, and has a context length of 262k.69% similarUnverifiedQwen3 27B uses linear attention in 48 of its 64 layers, with the remaining 16 layers paired with standard gated attention.66% similarUnverifiedOn the Artificial Analysis platform, Agnes-3.0-Flash is labeled as a Proprietary model with an AA composite score of 3665% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/916672API
curl https://kongchang.com/api/v1/knowledge/claims/916672MCP
get_claim(id=916672)