# Grok 2 API Rate Limits, Pricing & Performance (July 2026)

- Tool: AI API Limits & Performance Matrix
- Last updated: July 2026

## TL;DR

Grok 2 and Grok 2 Vision are retired legacy models with no current first-party xAI SKU promoted. Pricing for Grok 2 is no longer published. The recommended replacement is Grok 4.5, xAI's current flagship launched July 8, 2026: 500K-token context, $2.00 input / $6.00 output per MTok, cached input at $0.50/MTok, and 150 RPS / 50M TPM production limits. No free API tier.

## Frequently asked questions

### Does Grok 2 still work via the API?

Grok 2 / Grok 2 Vision is retired — xAI no longer promotes it as a current SKU. The slug may still resolve, but pricing is not published and the endpoint is not maintained. If your code references grok-2, migrate to the current Grok 4.5 API string (verify the exact name on docs.x.ai). Continuing to call a retired slug risks unexpected billing behavior or silent deprecation.

### What is the Grok 2 API price?

Grok 2 pricing is not published — the model is retired. Grok 2 Vision was historically $2.00 input / $10.00 output per MTok, but these rates no longer apply to any active endpoint. The current xAI flagship is Grok 4.5 at $2.00 input / $6.00 output per MTok with cached input at $0.50/MTok.

### How do I migrate from Grok 2 to Grok 4.5?

Update the model parameter in your API calls from 'grok-2' (or 'grok-2-vision') to the current Grok 4.5 string — verify the exact API name on docs.x.ai. The xAI API is OpenAI-SDK compatible, so no other client code changes are needed. Test against the new model in staging, as capabilities and token pricing differ. Grok 4.5 adds 500K context and reasoning modes not available in Grok 2.

### Is the xAI API free?

No. xAI ended its $150/month free credit program in May 2025. There is no free API tier. You must recharge your xAI account to access any Grok model. Billing for Grok 4.5 starts at $2.00/MTok input and $6.00/MTok output.

### How does Grok 4.5 compare to Mistral Large on pricing?

Grok 4.5 and Mistral Large 3 are at pricing parity on input ($2.00/MTok) and output ($6.00/MTok). Mistral has a 50% batch discount that Grok 4.5 lacks at launch. Grok 4.5 has a much larger 500K context window vs Mistral Large's 128K, and includes native X Search grounding. Mistral offers a free experimental tier and EU-based inference.

### What does HTTP 429 mean when calling the xAI API?

HTTP 429 Too Many Requests means you exceeded either the 150 RPS hard cap or the 50M TPM limit on Grok 4.5. The OpenAI SDK surfaces this as RateLimitError. xAI does not send a Retry-After header — implement exponential backoff (wait = 2 ** attempt seconds with random jitter). If you hit 429s calling grok-2, the fix is to migrate to grok-4.5 first, then tune concurrency.

### Can RapidDev help with migrating from Grok 2 to Grok 4.5 in production?

Yes. RapidDev handles model migration, gateway architecture, and quota engineering for production AI APIs. If you need help safely replacing Grok 2 with Grok 4.5 in an existing system — including handling rate limits and cost projections — book a free scoping call at rapidevelopers.com/contact.

---

Source: https://www.rapidevelopers.com/ai-api-limits-performance-matrix/grok-2
© RapidDev — https://www.rapidevelopers.com/ai-api-limits-performance-matrix/grok-2