Merge pull request #3914 from tgmerritt/pe/add_o3-mini

add support for o3-mini
continuedev · Feb 1, 2025 · cd97a73 · cd97a73
2 parents 9ae37ac + e7be878
commit cd97a73
Show file tree

Hide file tree

Showing 5 changed files with 31 additions and 6 deletions.
diff --git a/core/llm/llms/Asksage.ts b/core/llm/llms/Asksage.ts
@@ -24,6 +24,7 @@ class Asksage extends BaseLLM {
     "gpt4-32k": "gpt4-32k",
     "gpt-o1": "gpt-o1", // Works
     "gpt-o1-mini": "gpt-o1-mini", // Works
+    "gpt-o3-mini": "gpt-o3-mini", // Stub
     "gpt-3.5-turbo": "gpt35-16k", // Works
     "aws-bedrock-claude-35-sonnet-gov": "aws-bedrock-claude-35-sonnet-gov", // Works
     "claude-3-5-sonnet-latest": "claude-35-sonnet", // Works

diff --git a/core/llm/llms/OpenAI.ts b/core/llm/llms/OpenAI.ts
@@ -49,6 +49,7 @@ const CHAT_ONLY_MODELS = [
   "gpt-4o-mini",
   "o1-preview",
   "o1-mini",
+  "o3-mini"
 ];
 
 const formatMessageForO1 = (messages: ChatCompletionMessageParam[]) => {
@@ -92,8 +93,8 @@ class OpenAI extends BaseLLM {
     return model;
   }
 
-  private isO1Model(model?: string): boolean {
-    return !!model && model.startsWith("o1");
+  private isO3orO1Model(model?: string): boolean {
+    return !!model && (model.startsWith("o1") || model.startsWith("o3"));
   }
 
   protected supportsPrediction(model: string): boolean {
@@ -144,8 +145,8 @@ class OpenAI extends BaseLLM {
 
     finalOptions.stop = options.stop?.slice(0, this.getMaxStopWords());
 
-    // OpenAI o1-preview and o1-mini:
-    if (this.isO1Model(options.model)) {
+    // OpenAI o1-preview and o1-mini or o3-mini:
+    if (this.isO3orO1Model(options.model)) {
       // a) use max_completion_tokens instead of max_tokens
       finalOptions.max_completion_tokens = options.maxTokens;
       finalOptions.max_tokens = undefined;
@@ -239,8 +240,8 @@ class OpenAI extends BaseLLM {
   ): ChatCompletionCreateParams {
     body.stop = body.stop?.slice(0, this.getMaxStopWords());
 
-    // OpenAI o1-preview and o1-mini:
-    if (this.isO1Model(body.model)) {
+    // OpenAI o1-preview and o1-mini or o3-mini:
+    if (this.isO3orO1Model(body.model)) {
       // a) use max_completion_tokens instead of max_tokens
       body.max_completion_tokens = body.max_tokens;
       body.max_tokens = undefined;

diff --git a/docs/docs/customize/model-providers/more/asksage.md b/docs/docs/customize/model-providers/more/asksage.md
@@ -60,6 +60,7 @@ The current Models available provided by Ask Sage are:
 | 8     | GPT-4-32K              |  Yes  | ✅     |
 | 9     | GPT-o1                 |  Yes  | ✅     |
 | 10    | GPT-o1-mini            |  Yes  | ✅     |
+| 10    | GPT-o3-mini            |  Yes  | ✅     |
 | 11    | GPT-3.5-turbo          |  Yes  | ✅     |
 | 12    | Calude 3.5 Sonnet Gov  |  Yes  | ✅     |
 | 13    | Calude 3 Opus          |  Yes  | ✅     |

diff --git a/gui/src/pages/AddNewModel/configs/models.ts b/gui/src/pages/AddNewModel/configs/models.ts
@@ -1228,6 +1228,21 @@ export const models: { [key: string]: ModelPackage } = {
     icon: "openai.png",
     isOpenSource: false,
   },
+  asksagegpto3mini: {
+    title: "GPT-o3-mini",
+    description:
+      "o3-mini can outperform o1 in coding and other reasoning tasks, and is 93% cheaper and has lower latency. It supports function calling, Structured Outputs, streaming, and developer messages. o3-mini comes with a larger context window of 200,000 tokens and a max output of 100,000 tokens",
+    params: {
+      model: "gpt-o3-mini",
+      contextLength: 200_000,
+      title: "GPT-o3-mini",
+      systemMessage:
+        "You are an expert software developer. You give helpful and concise responses.",
+    },
+    providerOptions: ["askSage"],
+    icon: "openai.png",
+    isOpenSource: false,
+  },
   asksageclaude35gov: {
     title: "Claude 3.5 Sonnet gov",
     description:

diff --git a/packages/llm-info/src/providers/openai.ts b/packages/llm-info/src/providers/openai.ts
@@ -97,6 +97,13 @@ export const OpenAi: ModelProvider = {
       maxCompletionTokens: 65536,
       recommendedFor: ["chat"],
     },
+    {
+      model: "o3-mini",
+      displayName: "o3 Mini",
+      contextLength: 128000,
+      maxCompletionTokens: 65536,
+      recommendedFor: ["chat"],
+    },
     // embed
     {
       model: "text-embedding-3-large",