feat(google-vertex): update model YAMLs [bot]

[harshiv-26]

Auto-generated by poc-agent for provider google-vertex.

---

Note

Medium Risk
Primarily metadata updates, but changes to status/isDeprecated and supported modalities/features can alter which Vertex models are selectable and how requests are formed.

Overview
Updates Google Vertex model YAML metadata across Anthropic/Claude, Gemini, Imagen, DeepSeek, Llama, Mistral, Qwen, Moonshot, OpenAI, and Z.ai entries.

The diff adds/standardizes lifecycle fields (e.g., status: active|deprecated|preview, plus some isDeprecated and new deprecationDates) and adjusts declared capabilities (new/removed features, modality tweaks, embedding outputs, and a few parameter/cost field additions such as video token pricing and tool_use_system_prompt_tokens).

^{Written by Cursor Bugbot for commit e24fa5a. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

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

[cursor]

Output token limits removed from codestral-2 model

Medium Severity

The max_output_tokens and max_tokens fields were removed from the codestral-2 limits, leaving only context_window and max_input_tokens. The versioned variant codestral-2@001.yaml still retains both at 128000. This means the unversioned model no longer declares any output token limit, which could cause incorrect request validation or cost calculations for consumers of this data.

Additional Locations (1)

• providers/google-vertex/mistralai/codestral-2@001.yaml#L10-L15

 

[cursor]

Cache cost data removed but caching feature retained

Medium Severity

All cache_read_input_token_cost and cache_read_input_audio_token_cost fields were removed from every region's cost entry, but prompt_caching is still listed in features. Comparable models like gemini-2.5-flash retain both the feature and the corresponding cache cost fields. This inconsistency could cause incorrect cost calculations when prompt caching is used with this model.

Additional Locations (1)

• providers/google-vertex/gemini-3-flash-preview.yaml#L186-L187