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

[models-bot]

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

---

Note

Low Risk
Metadata-only YAML updates to modalities, docs links, and context limits; no runtime or auth logic changes.

Overview
Updates google-vertex model catalog YAML for two partner models.

For anthropic/claude-fable-5, PDF is now listed under input modalities, and a Claude effort documentation URL is added under sources (alongside existing pricing and Vertex docs).

For xai/grok-4.3, limits.context_window is raised from 200,000 to 1,000,000 tokens; features, modalities, and tiered pricing blocks are unchanged in this diff.

^{Reviewed by Cursor Bugbot for commit a41285c. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

/test-models

[harshiv-26]

Gateway test results

• Total: 18
• Passed: 18
• Failed: 0
• Validation failed: 0
• Errored: 0
• Skipped: 0
• Success rate: 100.0%

Provider	Model	Scenarios
google-vertex	anthropic/claude-fable-5	success: params, tool-call:stream, params:stream, tool-call, parallel-tool-call, structured-output:stream, structured-output, parallel-tool-call:stream, reasoning:stream, reasoning
google-vertex	xai/grok-4.3	success: tool-call:stream, tool-call, params, params:stream, structured-output:stream, structured-output, reasoning:stream, reasoning

Successes (18)

google-vertex/xai/grok-4.3 — tool-call:stream (success)

Output

{"location":"London"}
VALIDATION: tool-call stream SUCCESS

google-vertex/xai/grok-4.3 — tool-call (success)

Output

Function: get_weather
Arguments: {"location":"London"}
VALIDATION: tool-call SUCCESS

google-vertex/xai/grok-4.3 — params (success)

Output

The capital of France is **Paris**.

google-vertex/xai/grok-4.3 — params:stream (success)

Output

The capital of France is **Paris**.

google-vertex/xai/grok-4.3 — structured-output:stream (success)

Output

{
  "date": "Friday",
  "name": "science fair",
  "participants": ["Alice", "Bob"]
}
VALIDATION: structured-output stream SUCCESS

google-vertex/xai/grok-4.3 — structured-output (success)

Output

{
  "date": "Friday",
  "name": "Science Fair",
  "participants": ["Alice", "Bob"]
}
VALIDATION: structured-output SUCCESS

google-vertex/xai/grok-4.3 — reasoning:stream (success)

Output

**Step-by-step calculation of \(3^3^3^3\)**

Exponentiation is **right-associative**, so the expression \(3^3^3^3\) is interpreted as a power tower ev
... (truncated, 1041 chars omitted)

google-vertex/xai/grok-4.3 — reasoning (success)

Output

To evaluate the expression \(3^3^3^3\), first resolve the ambiguity in how the operators associate. Exponentiation is conventionally right-associative
... (truncated, 1701 chars omitted)

google-vertex/anthropic/claude-fable-5 — params (success)

Output

The capital of France is **Paris**.

google-vertex/anthropic/claude-fable-5 — tool-call:stream (success)

Output

{"location": "London"}
VALIDATION: tool-call stream SUCCESS

google-vertex/anthropic/claude-fable-5 — params:stream (success)

Output

The capital of France is **Paris**.

google-vertex/anthropic/claude-fable-5 — tool-call (success)

Output

Function: get_weather
Arguments: {"location":"London"}
VALIDATION: tool-call SUCCESS

google-vertex/anthropic/claude-fable-5 — parallel-tool-call (success)

Output

Number of parallel tool calls: 2
Function: get_weather
Arguments: {"location":"London"}
Function: get_weather
Arguments: {"location":"Paris"}
VALIDATI
... (truncated, 31 chars omitted)

google-vertex/anthropic/claude-fable-5 — structured-output:stream (success)

Output

{"name": "Science Fair", "date": "Friday", "participants": ["Alice","Bob"]}
VALIDATION: structured-output stream SUCCESS

google-vertex/anthropic/claude-fable-5 — structured-output (success)

Output

{"name":"Science Fair","date":"Friday","participants":["Alice","Bob"]}
VALIDATION: structured-output SUCCESS

google-vertex/anthropic/claude-fable-5 — parallel-tool-call:stream (success)

Output

I'll check the weather in both cities simultaneously:{"location": "London"}{"location": "Paris"}
Number of parallel tool calls: 2
VALIDATION: parallel
... (truncated, 26 chars omitted)

google-vertex/anthropic/claude-fable-5 — reasoning:stream (success)

Output

# Calculating 3^3^3^3

## Step 1: Understand the order of operations

Exponentiation is **right-associative**, meaning we evaluate from **top to botto
... (truncated, 1181 chars omitted)

google-vertex/anthropic/claude-fable-5 — reasoning (success)

Output

# Calculating 3^3^3^3

## Step 1: Understand the order of operations

Exponentiation is **right-associative**, meaning we evaluate from the top down (
... (truncated, 1322 chars omitted)

[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 a41285c. Configure here.}

[cursor]

Wrong Grok context window limit

Medium Severity

context_window for xai/grok-4.3 on Google Vertex was raised from 200000 to 1000000, but Google Cloud’s Grok 4.3 model card still lists a 200000-token context length. The 1M figure matches xAI’s general API, not this provider, so long-context callers can hit rejected requests.

 

^{Reviewed by Cursor Bugbot for commit a41285c. Configure here.}

[Archit-Kumar-16]

wrong change