[BUG] Batch API rejects grok-4.5, but Batch API docs examples use grok-4.5

[ljodea]

The Batch API rejects grok-4.5, but the Batch API docs use grok-4.5 in their examples.

Repro (docs example, verbatim)

from xai_sdk import Client
from xai_sdk.chat import user

client = Client(api_key=...)
batch = client.batch.create(batch_name="test")
chat = client.chat.create(model="grok-4.5", messages=[user("Say OK")], batch_request_id="t1")
client.batch.add(batch.batch_id, [chat])

grpc._channel._InactiveRpcError:
  status = StatusCode.INVALID_ARGUMENT
  details = "Model grok-4.5 is not supported for batch processing."

The REST equivalent (POST /v1/batches/{id}/requests) returns the same error with HTTP 400. Identical behavior from EU and US egress IPs (not the regional gating that affects grok-4.5 on chat completions). Same calls with grok-4.3 or grok-4.20-0309-* succeed.

Expected

Either batch support for grok-4.5, or docs examples using a model the Batch API accepts.

Environment: xai-sdk 1.17.0, Python 3.14, observed 2026-07-11 and 2026-07-12.

[Sammy-Dabbas]

I do not think this is fixable in the Python SDK. The SDK keeps no client-side allowlist of batch-eligible models, the model string you pass to client.chat.create is forwarded to the backend unchanged.

That is why the failure is a server-side gRPC INVALID_ARGUMENT and REST 400 saying "Model grok-4.5 is not supported for batch processing", it would otherwise be a Python error raised before the call. So it is a backend or docs fix on the xAI side, not the SDK.