docs: refresh benchmark scores (99.4% expanded quality)

README.md

@@ -30,12 +30,12 @@ Surgical refine on expanded-benchmark failures (browser commands, OOD retail/dev
 |------|----|----|---|
 | Core hard eval (117) | 100% | **100%** | — |
 | Core held-out (27) | 100% | **100%** | — |
-| Expanded hard (170) | 97.1% | **98.2%** | +1.1% |
+| Expanded hard quality (170) | 98.2% | **99.4%** | +1.2% |
 | Expanded held-out (39) | 97.4% | **100%** | +2.6% |
 | Generalization suite | 91.4% | **97.1%** | +5.7% |
-| p50 latency | ~1072ms | ~1066ms | ~same |
+| Core p50 latency | ~1028ms | **~1002ms** | ~same |
 
-**Zero regressions** vs v1 on the 170-case expanded set.
+Hard quality = task accuracy excluding latency SLA. v3 is the only variant that passes both production gates (100% core, ≥99% expanded quality).
 
 ## Quick start (MLX, Apple Silicon)