feat(xai): update model YAMLs [bot]

[models-bot]

Auto-generated by poc-agent for provider xai.

---

Note

Low Risk
Provider catalog YAML only; affects reported pricing and model availability metadata, not runtime application logic in this diff.

Overview
Updates xAI provider YAML definitions for several Grok models—mostly cost metadata and lifecycle status.

Tiered pricing (200k+ tokens, cumulative): Adds tiered_pricing to grok-4.20-experimental-beta-0304-non-reasoning (input/output) and grok-4.20-multi-agent-experimental-beta-0304 (cache read, input, output), matching the pattern already used on other Grok 4.x entries.

Batch pricing: Raises input_cost_per_token_batches and output_cost_per_token_batches on grok-4.20-multi-agent-latest (e.g. batch input from 6.25e-7 to 0.000001, batch output from 0.00000125 to 0.000002).

Status: Sets status: active on grok-4.5-latest and grok-imagine-image-2.0.

^{Reviewed by Cursor Bugbot for commit 7baa858. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

/test-models

[cursor]

Cursor Bugbot has reviewed your changes using default effort 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 7baa858. Configure here.}

[cursor]

Missing cache-read long-context tier

Medium Severity

The added tiered_pricing raises input and output after 200000 tokens but omits cache_read. This model has prompt_caching and a cache_read_input_token_cost, and official xAI rates plus sibling grok-4.20 non-reasoning YAMLs double cached input to 4e-7 past that threshold, so long-context cache reads stay underpriced.

 

^{Reviewed by Cursor Bugbot for commit 7baa858. Configure here.}