app.py

import gradio as gr

data = {
    "OpenAI GPT-4o": {"input": 2.50, "output": 10.00, "context": "128K", "speed": "80 tok/s"},
    "OpenAI GPT-4o-mini": {"input": 0.15, "output": 0.60, "context": "128K", "speed": "100 tok/s"},
    "OpenAI o1": {"input": 15.00, "output": 60.00, "context": "200K", "speed": "30 tok/s"},
    "OpenAI o3-mini": {"input": 1.10, "output": 4.40, "context": "200K", "speed": "60 tok/s"},
    "Anthropic Claude Sonnet 4": {"input": 3.00, "output": 15.00, "context": "200K", "speed": "70 tok/s"},
    "Anthropic Claude Haiku 3.5": {"input": 0.80, "output": 4.00, "context": "200K", "speed": "120 tok/s"},
    "Anthropic Claude Opus 4": {"input": 15.00, "output": 75.00, "context": "200K", "speed": "25 tok/s"},
    "Google Gemini 2.5 Pro": {"input": 1.25, "output": 10.00, "context": "1M", "speed": "60 tok/s"},
    "Google Gemini 2.5 Flash": {"input": 0.15, "output": 0.60, "context": "1M", "speed": "150 tok/s"},
    "Google Gemini 2.0 Flash": {"input": 0.10, "output": 0.40, "context": "1M", "speed": "180 tok/s"},
    "Meta Llama 3.3 70B": {"input": 0.18, "output": 0.18, "context": "131K", "speed": "90 tok/s"},
    "Meta Llama 4 Maverick": {"input": 0.20, "output": 0.20, "context": "1M", "speed": "100 tok/s"},
    "Mistral Large": {"input": 2.00, "output": 6.00, "context": "128K", "speed": "70 tok/s"},
    "Mistral Small": {"input": 0.10, "output": 0.30, "context": "128K", "speed": "130 tok/s"},
    "DeepSeek V3": {"input": 0.27, "output": 1.10, "context": "128K", "speed": "60 tok/s"},
    "DeepSeek R1": {"input": 0.55, "output": 2.19, "context": "128K", "speed": "30 tok/s"},
    "xAI Grok 2": {"input": 2.00, "output": 10.00, "context": "131K", "speed": "50 tok/s"},
    "xAI Grok 3 Mini": {"input": 0.30, "output": 0.50, "context": "131K", "speed": "120 tok/s"},
    "Groq Llama 3.3 70B": {"input": 0.05, "output": 0.05, "context": "131K", "speed": "300 tok/s"},
    "Cohere Command R+": {"input": 2.50, "output": 10.00, "context": "128K", "speed": "55 tok/s"},
}

model_names = list(data.keys())

def calculate_cost(model, input_tokens, output_tokens, requests_per_day):
    if model not in data:
        return "Select a model"
    
    d = data[model]
    input_cost = (input_tokens / 1_000_000) * d["input"]
    output_cost = (output_tokens / 1_000_000) * d["output"]
    per_request = input_cost + output_cost
    daily = per_request * requests_per_day
    monthly = daily * 30
    
    results = []
    for name, info in sorted(data.items(), key=lambda x: (input_tokens/1e6)*x[1]["input"] + (output_tokens/1e6)*x[1]["output"]):
        ic = (input_tokens / 1_000_000) * info["input"]
        oc = (output_tokens / 1_000_000) * info["output"]
        total = (ic + oc) * requests_per_day * 30
        marker = " << SELECTED" if name == model else ""
        results.append(f"| {name} | ${total:,.2f} | {info['context']} | {info['speed']} |{marker}")
    
    comparison = "\n".join(results)
    
    return f"""## {model}

| Metric | Value |
|--------|-------|
| Per request | ${per_request:.4f} |
| Daily ({requests_per_day} req) | ${daily:.2f} |
| **Monthly** | **${monthly:.2f}** |
| Context window | {d['context']} tokens |
| Speed | {d['speed']} |
| Input price | ${d['input']}/1M tokens |
| Output price | ${d['output']}/1M tokens |

---

## All Models (sorted by monthly cost)

| Model | Monthly Cost | Context | Speed |
|-------|-------------|---------|-------|
{comparison}

---

*Data from [ComparEdge](https://comparedge.com/best/llm) — Updated April 2026*
"""

with gr.Blocks(title="LLM API Cost Calculator 2026") as demo:
    gr.Markdown("# LLM API Cost Calculator 2026\nCompare costs across 20 LLM APIs. Data from [ComparEdge](https://comparedge.com/best/llm).")
    
    with gr.Row():
        model = gr.Dropdown(model_names, label="Model", value="OpenAI GPT-4o")
        input_tokens = gr.Slider(100, 100000, value=2000, step=100, label="Input tokens/request")
    with gr.Row():
        output_tokens = gr.Slider(100, 50000, value=1000, step=100, label="Output tokens/request")
        requests = gr.Slider(1, 10000, value=100, step=1, label="Requests/day")
    
    output = gr.Markdown(label="Cost Breakdown")
    
    for inp in [model, input_tokens, output_tokens, requests]:
        inp.change(calculate_cost, [model, input_tokens, output_tokens, requests], output)
    
    demo.load(calculate_cost, [model, input_tokens, output_tokens, requests], output)

if __name__ == "__main__":
    demo.launch()