Code execution

The Gemini API code execution feature enables the model to generate and run Python code and learn iteratively from the results until it arrives at a final output. You can use this code execution capability to build applications that benefit from code-based reasoning and that produce text output. For example, you could use code execution in an application that solves equations or processes text.

Code execution is available in both Vertex AI Studio and the Gemini API. In Vertex AI Studio, you can enable code execution under Advanced settings. The Gemini API provides code execution as a tool, similar to function calling. After you add code execution as a tool, the model decides when to use it.

The code execution environment includes the following libraries. You can't install your own libraries.

Supported models

Code execution is supported by the gemini-2.0-flash-exp model.

Get started with code execution

This section assumes that you've completed the setup and configuration steps shown in the Gemini API quickstart.

Enable code execution on the model

You can enable basic code execution as shown here:

REST

Before using any of the request data, make the following replacements:

  • GENERATE_RESPONSE_METHOD: The type of response that you want the model to generate. Choose a method that generates how you want the model's response to be returned:
    • streamGenerateContent: The response is streamed as it's being generated to reduce the perception of latency to a human audience.
    • generateContent: The response is returned after it's fully generated.
  • LOCATION: The region to process the request. Available options include the following:

    Click to expand a partial list of available regions

    • us-central1
    • us-west4
    • northamerica-northeast1
    • us-east4
    • us-west1
    • asia-northeast3
    • asia-southeast1
    • asia-northeast1
  • PROJECT_ID: Your project ID.
  • MODEL_ID: The model ID of the model that you want to use.
  • ROLE: The role in a conversation associated with the content. Specifying a role is required even in singleturn use cases. Acceptable values include the following:
    • USER: Specifies content that's sent by you.
    • MODEL: Specifies the model's response.
  • TEXT
     The text instructions to include in the prompt.

To send your request, choose one of these options:

curl

Save the request body in a file named request.json. Run the following command in the terminal to create or overwrite this file in the current directory: 

cat > request.json << 'EOF'
{
  "tools": [{'codeExecution': {}}],
  "contents": {
    "role": "ROLE",
    "parts": { "text": "TEXT" }
  },
}
EOF

Then execute the following command to send your REST request: 

curl -X POST \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://LOCATION-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/MODEL_ID:GENERATE_RESPONSE_METHOD"

PowerShell

Save the request body in a file named request.json. Run the following command in the terminal to create or overwrite this file in the current directory: 

@'
{
  "tools": [{'codeExecution': {}}],
  "contents": {
    "role": "ROLE",
    "parts": { "text": "TEXT" }
  },
}
'@  | Out-File -FilePath request.json -Encoding utf8

Then execute the following command to send your REST request: 

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
    -Method POST `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://LOCATION-aiplatform.googleapis.com/v1beta1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/MODEL_ID:GENERATE_RESPONSE_METHOD" | Select-Object -Expand Content

You should receive a JSON response similar to the following.

Python

from google import genai
from google.genai.types import Tool, ToolCodeExecution, GenerateContentConfig

client = genai.Client()
model_id = "gemini-2.0-flash-exp"

code_execution_tool = Tool(
    code_execution=ToolCodeExecution()
)
response = client.models.generate_content(
    model=model_id,
    contents="Calculate 20th fibonacci number. Then find the nearest palindrome to it.",
    config=GenerateContentConfig(
        tools=[code_execution_tool],
        temperature=0,
    ),
)
for part in response.candidates[0].content.parts:
    if part.executable_code:
        print(part.executable_code)
    if part.code_execution_result:
        print(part.code_execution_result)
# Example response:
# code='...' language='PYTHON'
# outcome='OUTCOME_OK' output='The 20th Fibonacci number is: 6765\n'
# code='...' language='PYTHON'
# outcome='OUTCOME_OK' output='Lower Palindrome: 6666\nHigher Palindrome: 6776\nNearest Palindrome to 6765: 6776\n'

Go

import (
  "bytes"
  "context"
  "flag"
  "fmt"
  "io"

  genai "google.golang.org/genai"
)

// codeExecution generates code for the given text prompt using Code Execution as a Tool.
func codeExecution(w io.Writer) error {
  modelName := "gemini-2.0-flash-exp"
  client, err := genai.NewClient(context.TODO(), &genai.ClientConfig{})
  if err != nil {
    return fmt.Errorf("NewClient: %w", err)
  }

  codeExecTool := genai.Tool{
    CodeExecution: &genai.ToolCodeExecution{},
  }
  config := &genai.GenerateContentConfig{
    Tools: []*genai.Tool{&codeExecTool},
  }

  textpart := genai.Text(`Calculate 20th fibonacci number. Then find the nearest palindrome to it.`)

  result, err := client.Models.GenerateContent(context.TODO(), modelName,
    &genai.ContentParts{textpart}, config)
  if err != nil {
    return fmt.Errorf("GenerateContent: %w", err)
  }

  for _, part := range result.Candidates[0].Content.Parts {
    if part.ExecutableCode != nil {
      fmt.Fprintf(w, "Code (%s):\n%s\n", part.ExecutableCode.Language, part.ExecutableCode.Code)
    }
    if part.CodeExecutionResult != nil {
      fmt.Fprintf(w, "Result (%s):\n %s\n", part.CodeExecutionResult.Outcome, part.CodeExecutionResult.Output)
    }
  }
  return nil
}

Use code execution in chat

You can also use code execution as part of a chat.

REST

curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json" \
https://us-central1-aiplatform.googleapis.com/v1beta1/projects/test-project/locations/us-central1/publishers/google/models/gemini-2.0-flash-exp:generateContent -d \
$'{
    "tools": [{'code_execution': {}}],
    "contents": [
      {
        "role": "user",
        "parts": {
          "text": "Can you print \"Hello world!\"?"
        }
      },
      {
        "role": "model",
        "parts": [
          {
            "text": ""
          },
          {
            "executable_code": {
              "language": "PYTHON",
              "code": "\nprint(\"hello world!\")\n"
            }
          },
          {
            "code_execution_result": {
              "outcome": "OUTCOME_OK",
              "output": "hello world!\n"
            }
          },
          {
            "text": "I have printed \"hello world!\" using the provided python code block. \n"
          }
        ],
      },
      {
        "role": "user",
        "parts": {
          "text": "What is the sum of the first 50 prime numbers? Generate and run code for the calculation, and make sure you get all 50."
        }
      }
    ]
  }'

Code execution versus function calling

Code execution and function calling are similar features:

  • Code execution lets the model run code in the API backend in a fixed, isolated environment.
  • Function calling lets you run the functions that the model requests, in whatever environment you want.

In general, you should prefer to use code execution if it can handle your use case. Code execution is simpler to use (you just enable it) and resolves in a single GenerateContent request, thus incurring a single charge. Function calling takes an additional GenerateContent request to send back the output from each function call, thus incurring multiple charges.

For most cases, you should use function calling if you have your own functions that you want to run locally, and you should use code execution if you'd like the API to write and run Python code for you and return the result.

Billing

There's no additional charge for enabling code execution from the Gemini API. You'll be billed at the current rate of input and output characters.

Here are a few other things to know about billing for code execution:

  • You're only billed once for the input tokens you pass to the model, and you're billed for the final output tokens returned to you by the model.
  • Tokens representing generated code are counted as output tokens.
  • Code execution results are also counted as output tokens.

Limitations

  • The model can only generate and execute code. It can't return other artifacts like media files.
  • The feature doesn't support file I/O or use cases that involve non-text output (for example, data plots or a CSV file upload).
  • Code execution can run for a maximum of 30 seconds before timing out.
  • In some cases, enabling code execution can lead to regressions in other areas of model output (for example, writing a story).