💸 What Does One Amazon Bedrock Prompt Cost? Find Out From the CLI (Hands-on)

How much does a single Amazon Bedrock prompt cost? To find out, you can use the AWS Command Line Interface (CLI) to send a prompt to Bedrock, analyze the response for token usage, and then convert those tokens into a monetary value. Here's a hands-on demonstration of the process:

1. **Set Up**: Ensure you have AWS CLI v2 installed and configured with the correct region (us-east-1) and permissions to access Bedrock. Verify you have access to the Amazon Nova Micro model.

2. **Send a Prompt**: Use the Converse API to send a simple prompt. Here's the command:

```

aws bedrock-runtime converse --region us-east-1 --model-id us.amazon.nova-micro-v1:0 --messages '[{"role":"user","content":[{"text":"Explain overfitting in one sentence."}]}]' --query '{"answer": output.message.content[0].text, "usage": usage}'

```

The response will include `inputTokens`, `outputTokens`, and `totalTokens` which represent the usage of the prompt.

3. **Calculate Costs**: From the `usage` block returned by the API, note the `inputTokens` and `outputTokens`. Multiply these by the respective prices per 1,000 tokens from Amazon Bedrock's pricing page. You can perform this calculation using a simple `awk` command in the terminal:

```

awk -v i=$INPUT_TOKENS -v o=$OUTPUT_TOKENS -v pi=$PRICE_PER_1000_INPUT_TOKENS -v po=$PRICE_PER_1000_OUTPUT_TOKENS 'BEGIN { printf "USD %.8f per prompt\n", i/1000*pi + o/1000*po }'

```

Replace `$INPUT_TOKENS`, `$OUTPUT_TOKENS`, `$PRICE_PER_1000_INPUT_TOKENS`, and `$PRICE_PER_1000_OUTPUT_TOKENS` with the actual values you obtained.

4. **Scale Up**: To estimate daily costs, adjust the prompt to request longer responses and observe how the token usage changes. You might also experiment with the `maxTokens` parameter to control the length of the response and its impact on cost.

5. **Cost Efficiency**: Notice that longer, more detailed outputs typically cost more due to the increased number of output tokens. You can also adjust the `maxTokens` to find a balance between response length and cost efficiency.

6. **Cleanup**: There's no need to delete any resources created by the Converse API; each call is stateless and does not leave behind any persistent resources.

In summary, the cost of an Amazon Bedrock prompt is determined by the number of input and output tokens it generates, with output tokens generally being more expensive. By using the AWS CLI to send prompts and analyzing the token usage, you can estimate the cost of each prompt and make informed decisions about prompt length and complexity to optimize your budget.