Hi LiteLLM maintainers,

Thank you for maintaining `model_prices_and_context_window.json` — it has become an important community reference for routing, pricing checks, and cost estimation.

While cross-auditing model metadata against primary provider sources, we found two Gemini entries that appear to diverge from Google's current published prices (https://ai.google.dev/gemini-api/docs/pricing, accessed 2026-07-04):

**1. `gemini-2.5-flash-preview-tts`** (also `gemini/gemini-2.5-flash-preview-tts`) - LiteLLM: `input_cost_per_token: 3e-07` ($0.30/M), `output_cost_per_token: 2.5e-06` ($2.50/M) - Google's pricing page: "Input price ... $0.50 (text)", "Output price ... $10.00 (audio)"

**2. `gemini-2.5-flash-native-audio-preview-12-2025`** (also the `gemini/` prefixed key) - LiteLLM: `input_cost_per_token: 3e-07` ($0.30/M), `output_cost_per_token: 2.5e-06` ($2.50/M) - Google's pricing page: "Input price ... $0.50 (text) $3.00 (audio / video)", "Output price ... $2.00 (text) $12.00 (audio)"

Context: we ran this audit while building FactQuire (https://factquire.com), a source-verified facts feed for commercial LLM APIs where every price/limit is tied to a verbatim quote from the primary…