Skip to content

llama-3.1-8b-instruct

Beta

Model ID: @cf/meta/llama-3.1-8b-instruct

The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models. The Llama 3.1 instruction tuned text only models are optimized for multilingual dialogue use cases and outperform many of the available open source and closed chat models on common industry benchmarks.

Terms & License

Properties

Task Type: Text Generation

Use the Playground

Try out this model with Workers AI Model Playground. It does not require any setup or authentication and an instant way to preview and test a model directly in the browser.

Launch the Model Playground

Worker - Streaming

export interface Env {
AI: Ai;
}
export default {
async fetch(request, env): Promise<Response> {
const messages = [
{ role: "system", content: "You are a friendly assistant" },
{
role: "user",
content: "What is the origin of the phrase Hello, World",
},
];
const stream = await env.AI.run("@cf/meta/llama-3.1-8b-instruct", {
messages,
stream: true,
});
return new Response(stream, {
headers: { "content-type": "text/event-stream" },
});
},
} satisfies ExportedHandler<Env>;

Worker

export interface Env {
AI: Ai;
}
export default {
async fetch(request, env): Promise<Response> {
const messages = [
{ role: "system", content: "You are a friendly assistant" },
{
role: "user",
content: "What is the origin of the phrase Hello, World",
},
];
const response = await env.AI.run("@cf/meta/llama-3.1-8b-instruct", { messages });
return Response.json(response);
},
} satisfies ExportedHandler<Env>;

Python

import os
import requests
ACCOUNT_ID = "your-account-id"
AUTH_TOKEN = os.environ.get("CLOUDFLARE_AUTH_TOKEN")
prompt = "Tell me all about PEP-8"
response = requests.post(
f"https://api.cloudflare.com/client/v4/accounts/{ACCOUNT_ID}/ai/run/@cf/meta/llama-3.1-8b-instruct",
headers={"Authorization": f"Bearer {AUTH_TOKEN}"},
json={
"messages": [
{"role": "system", "content": "You are a friendly assistant"},
{"role": "user", "content": prompt}
]
}
)
result = response.json()
print(result)

curl

Terminal window
curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run/@cf/meta/llama-3.1-8b-instruct \
-X POST \
-H "Authorization: Bearer $CLOUDFLARE_AUTH_TOKEN" \
-d '{ "messages": [{ "role": "system", "content": "You are a friendly assistant" }, { "role": "user", "content": "Why is pizza so good" }]}'

Prompting

Part of getting good results from text generation models is asking questions correctly. LLMs are usually trained with specific predefined templates, which should then be used with the model's tokenizer for better results when doing inference tasks

There are two ways to prompt text generation models with Workers AI:

Scoped prompts

This is the recommended method. With scoped prompts, Workers AI takes the burden of knowing and using different chat templates for different models and provides a unified interface to developers when building prompts and creating text generation tasks.

Scoped prompts are a list of messages. Each message defines two keys: the role and the content.

Typically, the role can be one of three options:

  • system - System messages define the AI's personality. You can use them to set rules and how you expect the AI to behave.

  • user - User messages are where you actually query the AI by providing a question or a conversation.

  • assistant - Assistant messages hint to the AI about the desired output format. Not all models support this role.

OpenAI has a good explanation of how they use these roles with their GPT models. Even though chat templates are flexible, other text generation models tend to follow the same conventions.

Here's an input example of a scoped prompt using system and user roles:

{
messages: [
{ role: "system", content: "you are a very funny comedian and you like emojis" },
{ role: "user", content: "tell me a joke about cloudflare" },
],
};

Here's a better example of a chat session using multiple iterations between the user and the assistant.

{
messages: [
{ role: "system", content: "you are a professional computer science assistant" },
{ role: "user", content: "what is WASM?" },
{ role: "assistant", content: "WASM (WebAssembly) is a binary instruction format that is designed to be a platform-agnostic" },
{ role: "user", content: "does Python compile to WASM?" },
{ role: "assistant", content: "No, Python does not directly compile to WebAssembly" },
{ role: "user", content: "what about Rust?" },
],
};

Note that different LLMs are trained with different templates for different use cases. While Workers AI tries its best to abstract the specifics of each LLM template from the developer through a unified API, you should always refer to the model documentation for details (we provide links in the table above.) For example, instruct models like Codellama are fine-tuned to respond to a user-provided instruction, while chat models expect fragments of dialogs as input.

Unscoped prompts

You can use unscoped prompts to send a single question to the model without worrying about providing any context. Workers AI will automatically convert your prompt input to a reasonable default scoped prompt internally so that you get the best possible prediction.

{
prompt: "tell me a joke about cloudflare";
}

You can also use unscoped prompts to construct the model chat template manually. In this case, you can use the raw parameter. Here's an input example of a Mistral chat template prompt:

{
prompt: "<s>[INST]comedian[/INST]</s>
[INST]tell me a joke about cloudflare[/INST]",
raw: true
};

Responses

Using streaming

The recommended method to handle text generation responses is streaming.

LLMs work internally by generating responses sequentially using a process of repeated inference — the full output of a LLM model is essentially a sequence of hundreds or thousands of individual prediction tasks. For this reason, while it only takes a few milliseconds to generate a single token, generating the full response takes longer, on the order of seconds.

You can use streaming to start displaying the response as soon as the first tokens are generated, and append each additional token until the response is complete. This yields a much better experience for the end user. Displaying text incrementally as it's generated not only provides instant responsiveness, but also gives the end-user time to read and interpret the text.

To enable, set the stream parameter to true.

Using the Workers API:

const stream = await env.AI.run('@cf/meta/llama-3.1-8b-instruct', {
stream: true,
messages,
});
return new Response(stream, {
headers: {
"content-type": "text/event-stream",
},
});

Using the REST API:

Terminal window
curl https://api.cloudflare.com/client/v4/accounts/{ACCOUNT_ID}/ai/run/@cf/meta/llama-3.1-8b-instruct \
-X POST \
-H "Authorization: Bearer {API_TOKEN}" \
-d '{ "stream": true, "messages": [{ "role": "system", "content": "You are a friendly assistant" }, { "role": "user", "content": "Why is pizza so good?" }]}'

Streaming responses use server-sent events; the are easy to use, simple to implement on the server side, standardized, and broadly available across many platforms natively or as a polyfill.

Terminal window
curl -X POST \
"https://api.cloudflare.com/client/v4/accounts/<account>/ai/run/@cf/meta/llama-3.1-8b-instruct" \
-H "Authorization: Bearer {API_TOKEN}" \
-H "Content-Type:application/json" \
-d '{ "prompt": "where is new york?", "stream": true }'
data: {"response":"New"}
data: {"response":" York"}
data: {"response":" is"}
data: {"response":" located"}
data: {"response":" in"}
data: {"response":" the"}
...
data: [DONE]

Handling streaming responses in the client

Below is an example showing how to parse this response in JavaScript, from the browser:

const source = new EventSource("/"); // Workers AI streaming endpoint
source.onmessage = (event) => {
if (event.data == "[DONE]") {
source.close();
return;
}
const data = JSON.parse(event.data);
el.innerHTML += data.response;
};

Non-streaming response

Non-streaming responses may be helpful in some contexts, and they are possible; however, be aware that we limit the maximum number of output sequence tokens to avoid timeouts. Whenever possible, use streaming.

{
"response": "The origin of the phrase \"Hello, World\" is not well-documented, but it is believed to have originated in the early days of computing. In the 1970s, when personal computers were first becoming popular, many programming languages, including C, had a simple \"Hello, World\" program that was used to demonstrate the basics of programming.\nThe idea behind the program was to print the words \"Hello, World\" on the screen, and it was often used as a first program for beginners to learn the basics of programming. Over time, the phrase \"Hello, World\" became a common greeting among programmers and computer enthusiasts, and it is now widely recognized as a symbol of the computing industry.\nIt's worth noting that the phrase \"Hello, World\" is not a specific phrase that was coined by any one person or organization, but rather a catchphrase that evolved over time as a result of its widespread use in the computing industry."
}

API Schema

The following schema is based on JSON Schema

Input JSON Schema

{
"type": "object",
"oneOf": [
{
"properties": {
"prompt": {
"type": "string",
"minLength": 1,
"maxLength": 131072
},
"raw": {
"type": "boolean",
"default": false
},
"stream": {
"type": "boolean",
"default": false
},
"max_tokens": {
"type": "integer",
"default": 256
},
"temperature": {
"type": "number",
"default": 0.6,
"minimum": 0,
"maximum": 5
},
"top_p": {
"type": "number",
"minimum": 0,
"maximum": 2
},
"top_k": {
"type": "integer",
"minimum": 1,
"maximum": 50
},
"seed": {
"type": "integer",
"minimum": 1,
"maximum": 9999999999
},
"repetition_penalty": {
"type": "number",
"minimum": 0,
"maximum": 2
},
"frequency_penalty": {
"type": "number",
"minimum": 0,
"maximum": 2
},
"presence_penalty": {
"type": "number",
"minimum": 0,
"maximum": 2
}
},
"required": [
"prompt"
]
},
{
"properties": {
"messages": {
"type": "array",
"items": {
"type": "object",
"properties": {
"role": {
"type": "string"
},
"content": {
"type": "string",
"maxLength": 131072
}
},
"required": [
"role",
"content"
]
}
},
"functions": {
"type": "array",
"items": {
"type": "object",
"properties": {
"name": {
"type": "string"
},
"code": {
"type": "string"
}
},
"required": [
"name",
"code"
]
}
},
"stream": {
"type": "boolean",
"default": false
},
"max_tokens": {
"type": "integer",
"default": 256
},
"temperature": {
"type": "number",
"default": 0.6,
"minimum": 0,
"maximum": 5
},
"top_p": {
"type": "number",
"minimum": 0,
"maximum": 2
},
"top_k": {
"type": "integer",
"minimum": 1,
"maximum": 50
},
"seed": {
"type": "integer",
"minimum": 1,
"maximum": 9999999999
},
"repetition_penalty": {
"type": "number",
"minimum": 0,
"maximum": 2
},
"frequency_penalty": {
"type": "number",
"minimum": 0,
"maximum": 2
},
"presence_penalty": {
"type": "number",
"minimum": 0,
"maximum": 2
}
},
"required": [
"messages"
]
}
]
}

Output JSON Schema

{
"oneOf": [
{
"type": "object",
"contentType": "application/json",
"properties": {
"response": {
"type": "string"
}
}
},
{
"type": "string",
"contentType": "text/event-stream",
"format": "binary"
}
]
}