diff --git a/README.md b/README.md
index 46a7b15..0d6fde8 100644
--- a/README.md
+++ b/README.md
@@ -93,21 +93,15 @@ cargo run --example synthesize --features qwen3-tts -- --model-dir /path/to/mode
 ## Performance
 
 <!-- bench:start -->
-| Backend | Precision | Provider | Hardware | RTF short | RTF medium | RTF long | Date |
-|---------|-----------|----------|----------|:-----------:|:-----------:|:-----------:|------|
-| qwen3-tts | int4 | CPU | Standard_NC4as_T4_v3 | 1.98 | 2.04 | 2.34 | 2026-04-07 |
-| qwen3-tts | int4 | CUDA | Standard_NC4as_T4_v3 | **0.78** | **0.85** | 1.07 | 2026-04-07 |
+| Backend | Precision | Provider | Hardware | RTF short | RTF medium | RTF long |
+|---------|-----------|----------|----------|:-----------:|:-----------:|:-----------:|
+| qwen3-tts | int4 | CPU | Standard_NC4as_T4_v3 | 1.98 | 2.04 | 2.34 |
+| qwen3-tts | int4 | CUDA | Standard_NC4as_T4_v3 | **0.78** | **0.85** | 1.07 |
 
 _RTF < 1.0 = faster-than-real-time. Lower is better._  
 _To update: run `make bench-csv-cuda` on target hardware, then commit `bench/results/`._
 <!-- bench:end -->
 
-## Try it on Google Colab
-
-No local GPU needed — run Qwen3-TTS on a free T4 in the browser:
-
-[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1qtc6lAk9RsAsvF1ojft0ACO2-PzFX4pi?usp=sharing)
-
 ## Feature flags
 
 ### Backends
@@ -121,11 +115,11 @@ No local GPU needed — run Qwen3-TTS on a free T4 in the browser:
 
 Composable with any backend flag. Selects the inference hardware at build time.
 
-| Flag | Description |
-|------|-------------|
-| `cuda` | NVIDIA CUDA GPU |
-| `tensorrt` | NVIDIA TensorRT |
-| `coreml` | Apple CoreML (macOS) |
+| Flag | Description | Status |
+|------|-------------|--------|
+| `cuda` | NVIDIA CUDA GPU | ✅ Working |
+| `tensorrt` | NVIDIA TensorRT | 🚧 Not configured |
+| `coreml` | Apple CoreML (macOS) | 🚧 Not configured |
 
 ## License