org card: G18b release (bench v25 39/39, 0 false done, threshold 0.96)

README.md

@@ -20,5 +20,7 @@ your own Mac, locally. 开源、本地运行的电脑操作 agent：看得清，
 | [brain-4b](https://huggingface.co/deskmind/brain-4b) | Strong tier: the steps the router escalates (MLX) |
 | [eyes-4b](https://huggingface.co/deskmind/eyes-4b) | Screen grounding: find the target on a screenshot |
 
-On the real macOS desktop (bench v23), the 0.8B → 4B router passes **92%** of tasks at **0.59 s/step** on an M4 Pro,
-with no false "done". It speaks the `/v1/systemone` API. Code: [github.com/deskmind-ai](https://github.com/deskmind-ai).
+Current Brain release: **G18b** (revision `g18b-q8`). On the real macOS desktop (bench v25, 13 tasks × 3 runs through
+the DeskMind app on an M4 Pro), the 0.8B → 4B router at threshold 0.96 passes **39/39** runs with no false "done".
+Decisions take about 0.5 s when the 0.8B answers and about 3.6 s when the 4B checks (about 70% of steps). The server is
+wire-compatible with `/v1/systemone`. Code: [github.com/deskmind-ai](https://github.com/deskmind-ai).