Grok API pricing by model and provider

Route-specific buying guide

How to budget Grok 4.6 on OpenRouter

Route ID: x-ai/grok-4.6

APIDir's accepted rows cover OpenRouter's Grok 4.6 base tier below 200,000 prompt tokens. They are not xAI direct prices and they exclude the route's higher long-context override.

OpenRouter reported a 500,000-token context window and a pricing override beginning at 200,000 prompt tokens when checked. Crossing that boundary can change token rates before the technical context limit is reached.

Cost formula

Use the displayed base rates only while prompt tokens remain below 200,000. For larger prompts, retrieve the live override first; then add separately metered tools such as web search and the cost of retries or fallback routes.

Costs that move the bill

• Whether prompt tokens stay below or cross the 200,000-token price boundary
• Completion length, cached-input eligibility, agent loops, and retry volume
• Separately metered web search or other tools enabled for the request
• Router fallback, regional availability, account limits, and production support

Checks before purchase

• Log prompt tokens before routing so the long-context tier cannot surprise the budget.
• Keep tool charges separate from input and output token estimates.
• Benchmark the exact x-ai/grok-4.6 route with realistic search and retry settings.
• Compare with xAI direct access only after route identity, threshold rules, and terms are aligned.

Source checked Aug 22, 2026: OpenRouter public model API. Recheck the live route before committing spend.

Estimate this route with the LLM workload calculator →

Compare another model family

• Qwen API pricing
• Kimi API pricing
• Claude API pricing

Decision guide

How to compare Grok API API providers

Grok API pricing belongs to an exact model and context band. This page currently compares Grok 4.6 through xAI's direct API and an OpenRouter route. The direct documentation has a higher long-context band, so the displayed standard rate is explicitly limited to requests below that threshold.

Price parity does not make the channels operationally identical. Routing behavior, supported endpoints, caching, quotas, region, data terms, support, and fallback controls can differ. Run the same representative workload and inspect the provider's final usage records.

What changes the Grok API API cost

• Exact Grok model, context threshold, prompt size, and output length
• Direct or router service, cached input, priority, and rate limits
• Tool calls, search features, retries, data retention, and operational support

Integration checks before production

• Reject requests that cross an unbudgeted context band
• Pin the provider route and disable unapproved fallback models
• Record token use, context band, route, retries, and provider request ID

Workloads this comparison can inform

Interactive reasoning

Output length and tool loops can dominate a matched input rate.

Large-context analysis

Use the long-context price band rather than extending the short-context estimate.

Router failover

Test model identity, error mapping, quota behavior, and data terms before production use.

How to read the price result

The two reviewed Grok 4.6 channels currently show matching short-context token meters. APIDir requires a third provider before a cheapest ranking and excludes the higher long-context band from this scope.

Related model price comparisons

• GPT API pricing comparisonCompare GPT API routes by exact model ID, context band, input, cached-input and output meters, provider channel, and source.
• Claude API pricing comparisonCompare Claude API routes by exact model, context tier, input and output meter, direct or router channel, source, and effective date.
• Qwen API pricing comparisonCompare Qwen API pricing by exact model ID, provider channel, input and output token meter, context, and official source.