ModelCap

Current AI model comparison

Exact identities · public evidence

Tencent

Hy4 preview

Position

#23

Index

71.8

Measured

9 reported rows against measured corpus ladders (41%); publisher-corpus-prior over 183 held-out anchors (11%); opens above measured predecessor tencent/hy3 on same-family ordering evidence (39%); 1 specialist-board observation at 2.5% support (9%); shrunk 1.3 toward the measured corpus

Comparison side 1

Tencent

TCAndon Router

Position

#138

Index

30.5

Inherited

finetune of Qwen/Qwen3-4B-Instruct-2507 · finetune (100%)

Comparison side 2

Output / 1M

$2.50

vs Unavailable

Context

1M

vs

Weights

Open weights

vs Open weights