Estimate Input Token Costs for LLM Gateway

1
import requests
2
import time
3
4
base_url = "https://api.assemblyai.com"
5
headers = {"authorization": "YOUR_API_KEY"}
6
7
# Transcribe audio file
8
audio_url = "https://assembly.ai/wildfires.mp3"
9
data = {"audio_url": audio_url}
10
11
response = requests.post(base_url + "/v2/transcript", headers=headers, json=data)
12
transcript_id = response.json()["id"]
13
polling_endpoint = base_url + f"/v2/transcript/{transcript_id}"
14
15
# Poll for completion
16
print("Waiting for transcription to complete...")
17
18
while True:
19
    transcript = requests.get(polling_endpoint, headers=headers).json()
20
    if transcript["status"] == "completed":
21
        break
22
    elif transcript["status"] == "error":
23
        raise RuntimeError(f"Transcription failed: {transcript['error']}")
24
    time.sleep(3)
25
26
# Define your prompt
27
prompt = "Provide a brief summary of the transcript."
28
29
# Calculate character count (transcript + prompt)
30
transcript_chars = len(transcript["text"])
31
prompt_chars = len(prompt)
32
total_chars = transcript_chars + prompt_chars
33
print(f"\nTotal characters: {total_chars}")
34
35
# Estimate tokens (roughly 4 characters = 1 token)
36
estimated_tokens = total_chars / 4
37
tokens_in_millions = estimated_tokens / 1_000_000
38
39
# Calculate input costs for different models
40
gpt5_cost = 1.25 * tokens_in_millions
41
claude_sonnet_cost = 3.00 * tokens_in_millions
42
gemini_pro_cost = 1.25 * tokens_in_millions
43
44
print(f"Estimated input tokens: {estimated_tokens:,.0f}")
45
print(f"\nEstimated input costs:")
46
print(f"GPT-5: ${gpt5_cost:.4f}")
47
print(f"Claude 4.5 Sonnet: ${claude_sonnet_cost:.4f}")
48
print(f"Gemini 2.5 Pro: ${gemini_pro_cost:.4f}")

1
import requests
2
import time
3
4
base_url = "https://api.assemblyai.com"
5
headers = {"authorization": "YOUR_API_KEY"}

1
audio_url = "https://assembly.ai/wildfires.mp3"
2
data = {"audio_url": audio_url}
3
4
response = requests.post(base_url + "/v2/transcript", headers=headers, json=data)
5
transcript_id = response.json()["id"]
6
polling_endpoint = base_url + f"/v2/transcript/{transcript_id}"
7
8
# Poll for completion
9
print("Waiting for transcription to complete...")
10
11
while True:
12
    transcript = requests.get(polling_endpoint, headers=headers).json()
13
    if transcript["status"] == "completed":
14
        break
15
    elif transcript["status"] == "error":
16
        raise RuntimeError(f"Transcription failed: {transcript['error']}")
17
    time.sleep(3)

1
# Define your prompt
2
prompt = "Provide a brief summary of the transcript."
3
4
# Calculate character count (transcript + prompt)
5
transcript_chars = len(transcript["text"])
6
prompt_chars = len(prompt)
7
total_chars = transcript_chars + prompt_chars
8
print(f"\nTotal characters: {total_chars}")

1
# Estimate tokens (roughly 4 characters = 1 token)
2
estimated_tokens = total_chars / 4
3
tokens_in_millions = estimated_tokens / 1_000_000
4
5
print(f"Estimated input tokens: {estimated_tokens:,.0f}")

1
# Calculate input costs for different models (rates per 1M tokens)
2
gpt5_cost = 1.25 * tokens_in_millions
3
claude_sonnet_cost = 3.00 * tokens_in_millions
4
gemini_pro_cost = 1.25 * tokens_in_millions
5
6
print(f"\nEstimated input costs:")
7
print(f"GPT-5: ${gpt5_cost:.4f}")
8
print(f"Claude 4.5 Sonnet: ${claude_sonnet_cost:.4f}")
9
print(f"Gemini 2.5 Pro: ${gemini_pro_cost:.4f}")