Grok 4.6 API pricing & cost
Current token rates, context limits, workload examples and unit-economics shortcuts for Grok 4.6.
Pricing page fallback verified September 3, 2026; client checks TokenCOGS live registry on load.
What the token rates mean in dollars
Example workload: 2,000 input tokens + 500 output tokens per request, standard text-token pricing.
| Volume | Estimated API cost | Approx. / request |
|---|---|---|
| 1,000 requests | $7.00 | $0.0070 |
| 100,000 requests | $700.00 | $0.0070 |
| 1,000,000 requests | $7,000.00 | $0.0070 |
Where Grok 4.6 sits on price
Grok 4.6 sits toward the lower-cost end of this provider’s current model lineup. Price alone is not a quality ranking, but this position is useful when you are screening models for unit-economics fit before benchmarking latency and quality.
Cost comparison is not a quality, latency or reliability ranking. Benchmark the task you actually need to run.
Compare all xAI models →What can move the invoice
- Cached input is tracked at $0.5 per 1M tokens, which can materially change repeated-system-prompt or repeated-context workloads.
- Long-context pricing rules apply above 200,000 input tokens in the current registry. Large-document and agent workloads should model that threshold explicitly.
Ways this provider may reduce spend
No verified provider-specific credits or promotions are currently stored in the TokenCOGS registry.
Turn pricing into unit economics
Grok 4.6 pricing questions
How much does Grok 4.6 cost per 1M tokens?
The current TokenCOGS snapshot lists Grok 4.6 at $2 per 1M input tokens and $6 per 1M output tokens. Live registry data can update these values when the provider changes pricing.
What would 1,000 Grok 4.6 requests cost?
At 2,000 input tokens and 500 output tokens per request, 1,000 requests are approximately $7.00 before provider-specific extras. Change the workload in the linked calculator for your own usage pattern.
Is Grok 4.6 the cheapest xAI model?
Grok 4.6 sits toward the lower-cost end of this provider’s current model lineup. Price alone is not a quality ranking, but this position is useful when you are screening models for unit-economics fit before benchmarking latency and quality.