Change supported models (#9271)

- add support for `gpt-4-1106-preview` and `gpt-3.5-turbo-1106`
- remove support for `text-davinci-003` and `text-davinci-002`

extensions/openai-gpt/CHANGELOG.md

@@ -1,5 +1,11 @@
 # AI Assistant Changelog
 
+## [Version: 1.6.0] - 2023-11-16
+
+- Add support for `gpt-4-1106-preview` and `gpt-3.5-turbo-1106` models
+- Remove support for `text-davinci-003` and `text-davinci-002` models
+- Update README.md
+
 ## [Version: 1.5.0] - 2023-03-30
 
 - Change name from `OpenAI GPT3` to `OpenAI GPT`

extensions/openai-gpt/README.md

@@ -33,7 +33,7 @@ The interface of the extension follows the interface of the OpenAI Playground.
 
 You can set different parameters for the AI model:
 
-`AI Model`: type of the model you want to use. `gpt-4` is the most powerful one for now, but `gpt-3.5-turbo` is cheaper, faster, and almost as capable.
+`AI Model`: type of the model you want to use. `gpt-4-1106-preview` is the most powerful one for now, but `gpt-3.5-turbo-1106` is cheaper, faster, and almost as capable.
 
 `Temperature`: controls randomness of the AI model. The lower it is, the less random (and "creative") the results will be.
 
@@ -47,10 +47,10 @@ You can set different parameters for the AI model:
 
 ### Supported AI Models
 
-1. `gpt-4`
-2. `gpt-3.5-turbo`
-3. `text-davinci-003`
-4. `text-davinci-002`
+1. `gpt-4-1106-preview`
+2. `gpt-3.5-turbo-1106`
+3. `gpt-4`
+4. `gpt-3.5-turbo`
 5. `text-curie-001`
 6. `text-babbage-001`
 7. `text-ada-001`

extensions/openai-gpt/src/ai.tsx

@@ -46,6 +46,8 @@ const configuration = new Configuration({
 const openai = new OpenAIApi(configuration);
 
 export default function Command() {
+  const maxTokensGPT41106Preview = 128000;
+  const maxTokensGPT35Turbo1106 = 16385;
   const maxTokensGPT4 = 8192;
   const maxTokensGPT35Turbo = 4096;
   const maxTokensDavinci = 4000;
@@ -64,10 +66,10 @@ export default function Command() {
   const [maxModelTokens, setMaxModelTokens] = useState<number>(maxTokensDavinci);
 
   const modelLimit = {} as modelTokenLimit;
+  modelLimit["gpt-4-1106-preview"] = maxTokensGPT41106Preview;
+  modelLimit["gpt-3.5-turbo-1106"] = maxTokensGPT35Turbo1106;
   modelLimit["gpt-4"] = maxTokensGPT4;
   modelLimit["gpt-3.5-turbo"] = maxTokensGPT35Turbo;
-  modelLimit["text-davinci-003"] = maxTokensDavinci;
-  modelLimit["text-davinci-002"] = maxTokensDavinci;
   modelLimit["text-curie-001"] = maxTokensAdaBabbageCurie;
   modelLimit["text-babbage-001"] = maxTokensAdaBabbageCurie;
   modelLimit["text-ada-001"] = maxTokensAdaBabbageCurie;
@@ -120,7 +122,10 @@ export default function Command() {
     setIsLoading(true);
     try {
       const completion: gptCompletion =
-        formRequest.model === "gpt-3.5-turbo" || formRequest.model === "gpt-4"
+        formRequest.model === "gpt-3.5-turbo" ||
+        formRequest.model === "gpt-4" ||
+        formRequest.model === "gpt-4-1106-preview" ||
+        formRequest.model === "gpt-3.5-turbo-1106"
           ? await openai.createChatCompletion({
               model: formRequest.model,
               messages: [
@@ -146,7 +151,10 @@ export default function Command() {
             });
       await showToast({ title: "Answer Received" });
       const response =
-        formRequest.model === "gpt-3.5-turbo" || formRequest.model === "gpt-4"
+        formRequest.model === "gpt-3.5-turbo" ||
+        formRequest.model === "gpt-4" ||
+        formRequest.model === "gpt-4-1106-preview" ||
+        formRequest.model === "gpt-3.5-turbo-1106"
           ? `\n\n${completion.data.choices[0].message.content}`
           : completion.data.choices[0].text;
       setTextPrompt(textPrompt + response);

extensions/openai-gpt/src/info-messages.ts

@@ -4,11 +4,11 @@ We generally recommend altering this or top_p but not both.
 
 Default: 0.7`;
 
-export const model = `The model which will generate the completion. Some models are more sutiable for certain tasks than others. "gpt-4" is the most powerful one, but "gpt-3.5-turbo" is cheaper, faster, and almost as capable`;
+export const model = `The model which will generate the completion. Some models are more sutiable for certain tasks than others. "gpt-4-1106-preview" is the most powerful one, but "gpt-3.5-turbo-1106" is cheaper, faster, and almost as capable`;
 
 export const maxTokens = `The maximum number of tokens to generate in the completion.
 
-The token count of your prompt plus this parameter cannot exceed the model's context length. "text-davinci-002" and "text-davinci-003" models have a context length of 4000 tokens, while the others have 2048.
+The token count of your prompt plus this parameter cannot exceed the model's context length. Please consult OpenAI documentation for the token limits.
 
 Default: 256`;