Add benchmark highlights and architecture figure to model card

README.md

@@ -20,6 +20,18 @@ time-series analysis together in one language model. Built on
 [Qwen3.6-27B](https://huggingface.co/Qwen/Qwen3.6-27B), it combines three LoRA
 experts through a learned controller that weights them for each request.
 
+A single TeeMoE checkpoint simultaneously ranks in the **top three on Context is
+Key, TimeSeriesExam, and GIFT-Eval** in the
+[paper's September 25, 2026 benchmark comparison](https://arxiv.org/html/2609.40265v1#S5.T1).
+
+| Benchmark | Metric | Score | Place |
+|---|---|---:|:---:|
+| Context is Key | RCRPS ↓ | **0.115** | **3rd** |
+| TimeSeriesExam v1.1 | Accuracy ↑ | **78.55%** | **1st** |
+| GIFT-Eval | Mean MASE rank ↓ | **19.990** | **3rd** |
+
+![Figure 1. OpenTSLM TeeMoE architecture: a learned controller composes three LoRA experts over a shared language backbone for forecasting and analysis.](https://raw.githubusercontent.com/OpenTSLM/OpenTSLM-TeeMoE/main/assets/teemoe_overview.png)
+
 ## Paper
 
 The model is described in **OpenTSLM TeeMoE: A Unified Time-Series Language Model