Allocate
Grok 4.5
APIGrok 4.5 is a language model from SpaceXAI with a 500K-token context window. Provider list price is $2 per million input tokens and $6 per million output; on Allocate you pay $2.14 and $6.42. It is a closed model served over API; the weights are not published.
Pricing
Prices checked 2026-07-21.
Price against its peers
Provider list prices per M tokens, Grok 4.5 against its nearest language peers by price.
What a real workload costs
Take 1,000,000 requests a month at 1,200 input and 350 output tokens each: 1,200M input and 350M output tokens. At list prices that is 1,200 × $2 + 350 × $6 = $4,500 a month. Billed on Allocate it is $4,815.
Grok 4.5 is served over API. Route traffic to it by name, meter every token, and swap it out in one click when a better fit ships.
Example usage
Point a route at xai/grok-4.5 and the endpoint never changes; swap the model behind it whenever you want.
curl https://api.allocate.network/v1/chat/completions \ -H "Authorization: Bearer $ALLOCATE_KEY" \ -d '{ "model": "xai/grok-4.5", "messages": [{"role": "user", "content": "Summarise the attached contract."}] }'
Common questions
How much does Grok 4.5 cost per million tokens?
Provider list price is $2 per million input tokens and $6 per million output tokens. On Allocate you pay $2.14 in and $6.42 out.
What context window does Grok 4.5 have?
500,000 tokens (500K). At roughly 0.75 words per token, that is about 375k words of English text per request.
What does cached input cost on Grok 4.5?
$0.30 per million tokens at list ($0.32 billed). Repeated prompt prefixes, such as a stable system prompt or tool definitions, bill at this rate instead of the full input price.
Can I fine-tune Grok 4.5?
No. Grok 4.5 is a closed model served over API; the weights are not published. If you want a model you can train and own, start from an open-weights base in the catalog and fine-tune that.
How do I call Grok 4.5 on Allocate?
Send xai/grok-4.5 in the model field of the OpenAI-compatible endpoint at api.allocate.network/v1, or point a route name (like prod/support-agent) at it so you can swap the model later without a deploy.