diff --git a/crates/ml/src/cuda_pipeline/c51_loss_kernel.cu b/crates/ml/src/cuda_pipeline/c51_loss_kernel.cu
index ec61ee726..449fa814d 100644
--- a/crates/ml/src/cuda_pipeline/c51_loss_kernel.cu
+++ b/crates/ml/src/cuda_pipeline/c51_loss_kernel.cu
@@ -497,7 +497,11 @@ extern "C" __global__ void c51_loss_batched(
                 min_eq = fminf(min_eq, eq_per_action[a]);
             }
             float q_gap_local = max_eq - min_eq;
-            float tau = fmaxf(q_gap_local, 0.01f);
+            /* Floor: proportional to mean Q magnitude so the Boltzmann is
+             * meaningful at any Q-value scale. 1% of |mean Q| or 1e-6. */
+            float mean_q = (max_eq + min_eq) * 0.5f;
+            float tau_floor = fmaxf(fabsf(mean_q) * 0.01f, 1e-6f);
+            float tau = fmaxf(q_gap_local, tau_floor);
             float sum_exp = 0.0f;
             for (int a = 0; a < n_d; a++) {
                 action_weights[a] = expf((eq_per_action[a] - max_eq) / tau);