[BUG] Improper Price for models with price differences based on input token count

[nickcatal]

Problem (one or two sentences)

The cost wrong for both Gemini and Sonnet 4/4.5 above 200k for every provider. And I believe Qwen 3 Max.

There is no calculator in the Roo codebase to reflect that an input token count over 200k is higher on Gemini and Claude than a prompt with an input token count under. Nor is there code in the Vertex AI provider to reflect regional differences on Sonnet 4.5 (once Sonnet 4/4.5 > 200k is supported)

The prices for Sonnet and Gemini baked into the code on Anthropic/Gemini/OpenRouter/Vertex/Bedrock/etc all assume that 200k-1m is the 0-200k price.

Context (who is affected and when)

All users of models with different prices when the input is above 200k tokens.

Reproduction steps

1. Use Gemini or Sonnet 4.5 on a provider that supports > 200k tokens, then compare your cost to the cost in Roo

Expected result

The correct price for these > 200k prices

Actual result

The same price as if the input is < 200k

Variations tried (optional)

No response

App Version

Latest

API Provider (optional)

None

Model Used (optional)

Claude 4/4.5, Gemini, (maybe) Qwen 3 Max

Roo Code Task Links (optional)

No response

Relevant logs or errors (optional)

[roomote]

I'm exploring a potential fix for the pricing calculation issue with models that have different rates for input tokens over 200k. I'll open a PR to share progress.

[roomote]

Opened a PR with my attempt at addressing this issue: #8984

The PR implements the core tiered pricing engine with comprehensive tests. The pricing calculation functions now support models with different rates based on input token count thresholds.

What's done:

• ✅ Core pricing engine that selects rates based on total input tokens
• ✅ Support for both Anthropic and OpenAI pricing models
• ✅ Comprehensive test coverage

What needs follow-up:

• Model registry updates to add tier configurations for Gemini, Claude Sonnet 4/4.5, and other affected models
• Vertex AI regional pricing implementation
• Provider-level integration tests

The foundation is in place - once model configurations are updated with tier definitions, the pricing calculations will automatically use the correct rates. Feedback welcome!