List your company

Language Modeling on WikiText v1 (test)

13.33PerplexityFusedKV
12.508418.054223.629.1458Nov 21, 2025Nov 23, 2025Nov 25, 2025Nov 27, 2025Nov 29, 2025Dec 1, 2025Dec 3, 2025

Evaluation results

MethodLinks
FusedKVModel Scale=1.5B2025.1213.33
FusedKV-LiteModel Scale=1.5B2025.1213.45
YOCOModel Scale=1.5B2025.1213.65
VanillaModel Scale=1.5B, Cache Mem.=12025.1213.67
GQAModel Scale=1.5B2025.1213.74
CLAModel Scale=1.5B, Cache Mem.=1/22025.1214.19
Selective RoPEModel Architecture=GLA, Parameters=1.3B, Training Tokens=26B2025.1117.87
FusedKVModel Scale=650M2025.1218.09
NoPEModel Architecture=GLA, Parameters=1.3B, Training Tokens=26B2025.1118.18
VanillaModel Scale=650M, Cache Mem.=12025.1218.47
RoPEModel Architecture=GLA, Parameters=1.3B, Training Tokens=26B2025.1118.5
FusedKV-LiteModel Scale=650M2025.1218.55
GQAModel Scale=650M2025.1219.05
YOCOModel Scale=650M2025.1219.21
CLAModel Scale=650M, Cache Mem.=1/22025.1219.6
FusedKVModel Scale=332M2025.1222.35
FusedKV-LiteModel Scale=332M2025.1222.78
VanillaModel Scale=332M, Cache Mem.=12025.1222.85
GQAModel Scale=332M2025.1223.29
YOCOModel Scale=332M2025.1223.31
CLAModel Scale=332M, Cache Mem.=1/22025.1224.62
RoPEModel Architecture=FoX, Parameters=370M, Training Tokens=10B, Output-norm setting=OFF2025.1125.29
NoPEModel Architecture=FoX, Parameters=370M, Training Tokens=10B, Output-norm setting=OFF2025.1125.52
NoPEModel Architecture=GLA, Parameters=370M, Training Tokens=10B2025.1125.72
NoPEModel Architecture=Gated DeltaNet, Parameters=~400M, Training Tokens=10B2025.1125.82
RoPEModel Architecture=Gated DeltaNet, Parameters=~400M, Training Tokens=10B2025.1125.92
Selective RoPEModel Architecture=GLA, Parameters=370M, Training Tokens=10B2025.1125.98
Selective RoPEModel Architecture=Gated DeltaNet, Parameters=~400M, Training Tokens=10B, Variant=stable_selective_rope2025.1126.13
RoPEModel Architecture=GLA, Parameters=370M, Training Tokens=10B2025.1126.51
Selective RoPEModel Architecture=FoX, Parameters=370M, Training Tokens=10B, Output-norm setting=OFF2025.1133.87