DeepSeek-V4 uses SiluAndMulWithClamp(10.0) which clamps: - silu(gate) to max 10.0 - up to [-10.0, 10.0] Our runner was doing plain F.silu(gate) * up without clamping. Large gate values could produce unbounded SiLU output, causing numerical issues in the L2 GEMM. This is likely contributing to garbage model output.