feat(openrouter): update model YAMLs [bot]

[models-bot]

Auto-generated by poc-agent for provider openrouter.

---

Note

Low Risk
Low risk config-only change that updates pricing metadata for x-ai/grok-4.20; main risk is incorrect cost reporting/billing estimates if the new rates are wrong.

Overview
Updates providers/openrouter/x-ai/grok-4.20.yaml pricing by lowering input_cost_per_token/output_cost_per_token and removing cache_read_input_token_cost plus the tiered_pricing overrides, leaving a single flat rate for region "*".

^{Reviewed by Cursor Bugbot for commit 2ac657e. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

/test-models

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 2ac657e. Configure here.}

[cursor]

Missing cache read cost despite prompt_caching feature listed

Medium Severity

The cache_read_input_token_cost field was removed from the costs section, but prompt_caching is still listed as a feature. Every other model in the OpenRouter provider that advertises prompt_caching (e.g., grok-4.yaml, grok-4.20-multi-agent.yaml, aion-2.0.yaml, nova-premier-v1.yaml) includes a corresponding cache_read_input_token_cost. This inconsistency could cause incorrect cost calculations for cached token reads.

 

^{Reviewed by Cursor Bugbot for commit 2ac657e. Configure here.}

[harshiv-26]

Gateway test results

• Total: 10
• Passed: 10
• Failed: 0
• Validation failed: 0
• Errored: 0
• Skipped: 0
• Success rate: 100.0%

Provider	Model	Scenarios
openrouter	x-ai/grok-4.20	success: params:text-text, tool-call:text-text, params:text-text:stream, tool-call:text-text:stream, structured-output:text-text:stream, structured-output:text-text, json-output:text-text:stream, json-output:text-text, reasoning:text-text:stream, reasoning:text-text