Add Mistral Small 3 model (#3308)

stanford-crfm · Feb 3, 2025 · 2401e5e · 2401e5e
1 parent 228e0f1
commit 2401e5e
Show file tree

Hide file tree

Showing 3 changed files with 22 additions and 0 deletions.
diff --git a/src/helm/config/model_deployments.yaml b/src/helm/config/model_deployments.yaml
@@ -1666,6 +1666,13 @@ model_deployments:
     client_spec:
       class_name: "helm.clients.mistral_client.MistralAIClient"
 
+  - name: mistralai/mistral-small-2501
+    model_name: mistralai/mistral-small-2501
+    tokenizer_name: mistralai/Mistral-Small-24B-Instruct-2501
+    max_sequence_length: 32000
+    client_spec:
+      class_name: "helm.clients.mistral_client.MistralAIClient"
+
   - name: mistralai/mistral-medium-2312
     model_name: mistralai/mistral-medium-2312
     tokenizer_name: mistralai/Mistral-7B-v0.1

diff --git a/src/helm/config/model_metadata.yaml b/src/helm/config/model_metadata.yaml
@@ -2134,6 +2134,15 @@ models:
     release_date: 2024-09-18
     tags: [TEXT_MODEL_TAG, LIMITED_FUNCTIONALITY_TEXT_MODEL_TAG, INSTRUCTION_FOLLOWING_MODEL_TAG]
 
+  - name: mistralai/mistral-small-2501
+    display_name: Mistral Small 3 (2501)
+    description: Mistral Small 3 (2501) is a pre-trained and instructed model catered to the '80%' of generative AI tasks—those that require robust language and instruction following performance, with very low latency. ([blog](https://mistral.ai/news/mistral-small-3/))
+    creator_organization_name: Mistral AI
+    access: open
+    num_parameters: 23600000000
+    release_date: 2025-01-30
+    tags: [TEXT_MODEL_TAG, LIMITED_FUNCTIONALITY_TEXT_MODEL_TAG, INSTRUCTION_FOLLOWING_MODEL_TAG]
+
   - name: mistralai/mistral-medium-2312
     display_name: Mistral Medium (2312)
     description: Mistral is a transformer model that uses Grouped-Query Attention (GQA) and Sliding-Window Attention (SWA).

diff --git a/src/helm/config/tokenizer_configs.yaml b/src/helm/config/tokenizer_configs.yaml
@@ -484,6 +484,12 @@ tokenizer_configs:
     end_of_text_token: "</s>"
     prefix_token: "<s>"
 
+  - name: mistralai/Mistral-Small-24B-Instruct-2501
+    tokenizer_spec:
+      class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
+    end_of_text_token: "</s>"
+    prefix_token: "<s>"
+
   # Nectec
   - name: nectec/OpenThaiLLM-Prebuilt-7B
     tokenizer_spec: