fix(gemini): include IMAGE tokens in input cost calculation

[Chesars]

Relevant issues

Fixes #14285

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem

Type

🐛 Bug Fix

Changes

When sending images as input to Gemini, LiteLLM was not counting IMAGE tokens for cost calculation. The API returns promptTokensDetails with separate modalities:

{
  "promptTokenCount": 262,
  "promptTokensDetails": [
    {"modality": "TEXT", "tokenCount": 4},
    {"modality": "IMAGE", "tokenCount": 258}
  ]
}

LiteLLM only counted TEXT tokens (4), ignoring IMAGE tokens (258), resulting in ~98% undercharged costs for image inputs.

Example

response = litellm.completion(
    model="gemini/gemini-2.0-flash",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "What is this?"},
            {"type": "image_url", "image_url": {"url": "https://example.com/image.jpg"}}
        ]
    }]
)
# Before fix: text_tokens=4, cost based on 4 tokens
# After fix: text_tokens=262, cost based on 262 tokens

Solution

According to Google's Gemini API pricing, image input tokens are priced the same as text tokens for all Gemini models:

Input price: $0.30 (text / image / video)

The fix adds IMAGE tokens to text_tokens in _calculate_usage():

elif detail["modality"] in ("TEXT", "IMAGE"):
    # Image tokens are priced the same as text tokens for Gemini
    text_tokens = (text_tokens or 0) + detail.get("tokenCount", 0)

Tests Added

• test_vertex_ai_usage_metadata_with_image_input_tokens - IMAGE tokens in promptTokensDetails
• test_vertex_ai_usage_metadata_with_image_input_tokens_reverse_order - Order independence

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Review	Updated (UTC)
litellm	Ready	Preview, Comment	Dec 17, 2025 11:26pm