The Berkeley Function Calling Leaderboard V3 (also called Berkeley Tool Calling Leaderboard V3) evaluates the LLM's ability to call functions (aka tools) accurately. This leaderboard consists of real-world data and will be updated periodically. For more information on the evaluation dataset and methodology, please refer to our blogs: BFCL-v1 introducing AST as an evaluation metric, BFCL-v2 introducing enterprise and OSS-contributed functions, and BFCL-v3 introducing multi-turn interactions. Checkout code and data.

As of today: 27/03/2025 the chart leaders are:

This fantastic resource is very valuable for all AI Researchers.

Here is a simple C# Code sample:

/// <summary>
/// Test the LLM with tools.
/// </summary>
private async void TestLLMWithTools()
{

    // Init a new Tool Registry:
    ToolRegistry toolRegistry = new ToolRegistry();

    // Init a new LlmClient:
    LlmClient llmClient = new LlmClient("http://192.168.0.2:11434", "watt-tool-8B", toolRegistry);

    List<string> queries = new List<string>();

    // search_the_web questions
    queries.Add("What are the top 5 AI companies in 2025?");
    queries.Add("Search for recent breakthroughs in quantum computing");
    queries.Add("Find me information about Mars colonization plans");

    // get_ip_address questions
    queries.Add("What's the keey words in the sentence: 'Send an email to john@example.com with subject 'Meeting' and body 'Let's meet tomorrow'");
    queries.Add("What are the key words in: Can you tell me the system's public IP?");
    queries.Add("What key words can you find in: Please show me my IP address");

    // get_ip_address questions
    queries.Add("What's my current IP address?");
    queries.Add("Can you tell me the system's public IP?");
    queries.Add("Please show me my IP address");

    // send_email questions
    queries.Add("Send an email to john@example.com with subject 'Meeting' and body 'Let's meet tomorrow'");
    queries.Add("Email sarah@test.com with subject 'Report' and body 'Here is the latest report'");
    queries.Add("Please send an email to mike@work.com, subject 'Update', body 'Project is on track'");

    // financial_ratios.interest_coverage questions
    queries.Add("What's Tesla's interest coverage ratio for the past 3 years?");
    queries.Add("Calculate Apple's interest coverage ratio over 5 years");
    queries.Add("Show me Microsoft’s interest coverage ratio for the last 2 years");

    // sales_growth.calculate questions
    queries.Add("What’s the sales growth rate for Amazon over the past 4 years?");
    queries.Add("Calculate Google's sales growth for the last 3 years");
    queries.Add("Find the sales growth rate of Netflix for the past 5 years");

    // weather_forecast questions
    queries.Add("What's the weather forecast for New York for the next 3 days?");
    queries.Add("Tell me the 5-day weather forecast for London");
    queries.Add("Show me the weather prediction for Tokyo for the next 2 days");

    // get_current_time questions
    queries.Add("What's the current time in ISO format?");
    queries.Add("Show me the time in Unix format");
    queries.Add("What’s the present time in ISO format?");


    // 
    foreach (string query in queries)
    {

        // 
        string json = await llmClient.CallLLMWithToolsAsync("/api/chat", query);

        // 
        Response response = JsonSerializer.Deserialize<Response>(json);
        textBox1.Text = response.Message.Content;
    }
}

This small test gives me a 100% success rate, which is fantastic for a small model like this!

Note: Ollama Tool Calls are not supported by the Watt Models, however, the models can be used like standard chat, and the function calling works as if you're chatting with a model.