feat: add Nemotron-Cascade-14B-Thinking model configuration to catalogue and README

modal/README.md

@@ -15,7 +15,7 @@ modal/
   service.py        Reusable serving layer: image + vllm command, register_model()
                     (provider-agnostic). Imports ModelConfig from catalogue.
   registry.py       Back-compat re-export of the catalogue's model lists.
-  app_nvidia.py     App "nvidia-llms"  — Nemotron 3 Nano 4B + 30B.
+  app_nvidia.py     App "nvidia-llms"  — Nemotron 3 Nano 4B + 30B, Cascade 14B Thinking.
   app_openbmb.py    App "openbmb-llms" — MiniCPM4.1-8B + MiniCPM-o 4.5.
   app_google.py     App "google-llms"  — Gemma 4 12B + 26B.
   client.py         OpenAI-SDK smoke-test client for any endpoint.
@@ -36,6 +36,7 @@ Each running endpoint also self-documents at `/docs` (Swagger UI) and
 | Provider | App            | Model                                   | Endpoint name         | GPU     |
 | -------- | -------------- | --------------------------------------- | --------------------- | ------- |
 | NVIDIA   | `nvidia-llms`  | NVIDIA-Nemotron-3-Nano-30B-A3B-BF16     | `nemotron-3-nano-30b` | H200:1  |
+| NVIDIA   | `nvidia-llms`  | Nemotron-Cascade-14B-Thinking           | `nemotron-cascade-14b-thinking` | L40S:1  |
 | NVIDIA   | `nvidia-llms`  | NVIDIA-Nemotron-3-Nano-4B-BF16          | `nemotron-3-nano-4b`  | L4:1    |
 | OpenBMB  | `openbmb-llms` | MiniCPM-o-4_5 (omni)                    | `minicpm-o-4-5`       | L40S:1  |
 | OpenBMB  | `openbmb-llms` | MiniCPM4.1-8B                           | `minicpm-4-1-8b`      | L40S:1  |

modal/catalogue.py

@@ -139,6 +139,23 @@ NVIDIA_MODELS: tuple[ModelConfig, ...] = (
         gated=True,
         max_concurrent_inputs=64,
     ),
+    ModelConfig(
+        name="nvidia/Nemotron-Cascade-14B-Thinking",
+        endpoint_name="nemotron-cascade-14b-thinking",
+        # Dense 14B reasoning model built on Qwen3-14B Base; thinking-only. BF16
+        # weights (~28GB) plus KV cache fit a single 48GB L40S. A specialist
+        # model — left unbound so it can be cast explicitly at a reasoning-heavy
+        # agent (e.g. the Judge) without displacing a tier default.
+        params_b=14,
+        gpu="L40S:1",
+        max_model_len=32768,
+        # Qwen3-native in vLLM (no custom code); ChatML template with a thinking
+        # block parsed by the Qwen3 reasoning parser.
+        reasoning_parser="qwen3",
+        tool_call_parser="hermes",
+        enable_auto_tool_choice=True,
+        max_concurrent_inputs=48,
+    ),
 )
 
 # --- OpenBMB (MiniCPM) ---------------------------------------------------------