Add Gemini 3.1 Pro (#4076)

Author: yifanmai

src/helm/config/model_deployments.yaml

@@ -1297,6 +1297,17 @@ model_deployments:
         thinking_config:
           include_thoughts: true
 
+  - name: google/gemini-3.1-pro-preview
+    model_name: google/gemini-3.1-pro-preview
+    tokenizer_name: google/gemma-2b  # Gemini has no tokenizer endpoint, so we approximate by using Gemma's tokenizer.
+    max_sequence_length: 1048576  # Source: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-1-pro
+    # TODO: Max output tokens: 65536
+    client_spec:
+      class_name: "helm.clients.google_genai_client.GoogleGenAIClient"
+      args:
+        thinking_config:
+          include_thoughts: true
+
   - name: google/gemini-robotics-er-1.5-preview
     model_name: google/gemini-robotics-er-1.5-preview
     tokenizer_name: google/gemma-2b  # Gemini has no tokenizer endpoint, so we approximate by using Gemma's tokenizer.

src/helm/config/model_metadata.yaml

@@ -1449,6 +1449,14 @@ models:
     release_date: 2025-11-18
     tags: [TEXT_MODEL_TAG, VISION_LANGUAGE_MODEL_TAG, AUDIO_LANGUAGE_MODEL_TAG, GOOGLE_GEMINI_MODEL_TAG, LIMITED_FUNCTIONALITY_TEXT_MODEL_TAG, INSTRUCTION_FOLLOWING_MODEL_TAG]
 
+  - name: google/gemini-3.1-pro-preview
+    display_name: Gemini 3.1 Pro (Preview)
+    description: Gemini 3.1 Pro is the next iteration in the Gemini 3 series of models, a suite of highly capable, natively multimodal reasoning models. ([blog](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/), [model card](https://deepmind.google/models/model-cards/gemini-3-1-pro/))
+    creator_organization_name: Google
+    access: limited
+    release_date: 2026-02-19
+    tags: [TEXT_MODEL_TAG, VISION_LANGUAGE_MODEL_TAG, AUDIO_LANGUAGE_MODEL_TAG, GOOGLE_GEMINI_MODEL_TAG, LIMITED_FUNCTIONALITY_TEXT_MODEL_TAG, INSTRUCTION_FOLLOWING_MODEL_TAG]
+
   - name: google/gemini-robotics-er-1.5-preview
     display_name: Gemini Robotics-ER 1.5
     description: Gemini Robotics-ER 1.5 is a vision-language model (VLM) designed for advanced reasoning in the physical world, allowing robots to interpret complex visual data, perform spatial reasoning, and plan actions from natural language commands.