large: ubatch 4096 (match v3), add --no-mmap; sync bench harness prod config

scripts/frank-default-large.env

@@ -51,9 +51,10 @@ LISTEN_PORT=52625
 CTX_SIZE=262144
 PARALLEL=1
 # Local override: the pinned runner uses 2048/512; production chat uses a
-# larger 4096/2048 batch (unified memory has headroom).
+# larger 4096/4096 batch (unified memory has headroom; matches the retired
+# v3 profile).
 BATCH_SIZE=4096
-UBATCH_SIZE=2048
+UBATCH_SIZE=4096
 TARGET_THREADS=16
 BATCH_THREADS=32
 CACHE_TYPE_K=f16