Every Grok AI Model Is Now Available Through Cloudflare's AI Gateway

The complete Grok model lineup on AI Gateway

Cloudflare's AI Gateway now surfaces every Grok model xAI currently offers through its API :

• Grok 4.3 — xAI's flagship text-and-image reasoning model built around a 1-million-token context window. It supports native tool calling, function calling, structured outputs, and configurable reasoning effort levels (none, low, medium, high). Accepts both text and image inputs .
• Grok Build 0.1 — A fast coding model trained for agentic coding workflows, currently in beta. It carries a 256,000-token context window and is priced lower than the reasoning flagship .
• Grok Imagine — Text-to-image generation models, including a higher-fidelity "Image Quality" variant that defaults to 2K output resolution and supports editing via reference images and masks .
• Grok Aurora — A video generation model that creates clips from text prompts and still images, with motion and audio generation .
• Grok Voice (STT) — A speech-to-text model supporting 25 languages with word-level timestamps, multichannel transcription, speaker diarization, and key-term biasing .
• Grok 4.20 — The previous flagship model, also accessible through the gateway for teams that haven't yet migrated .

Pricing for the text models

Pricing through Cloudflare's AI Gateway mirrors xAI's standard API rates. The two primary text models cost :

Relative to Grok 4.20 — which was priced at $2.00 per million input tokens and $6.00 per million output tokens — Grok 4.3 represents roughly a 37% reduction on input and a 58% reduction on output . This price cut, which took effect when xAI shipped Grok 4.3's general availability on May 6, 2026, is one of the more aggressive pricing moves in the frontier model market .

Billing flows through Cloudflare's unified billing system. Developers purchase credits in the Cloudflare Dashboard and spend them across any connected AI provider, including xAI. This eliminates the need to maintain separate billing relationships with each model vendor .

How developers route requests

Routing Grok requests through AI Gateway requires a straightforward URL swap. Developers replace the standard xAI API base URL with the Cloudflare gateway endpoint :

Three pieces of information are needed: the Cloudflare account ID, the gateway name configured in the dashboard, and a valid xAI API key. Because the gateway exposes an OpenAI-compatible API schema, existing integrations built on the OpenAI SDK work without code changes beyond the base URL update .

What Cloudflare's AI Gateway provides

Cloudflare positions AI Gateway as a control plane that sits between applications and model providers. Once Grok traffic flows through it, developers unlock several infrastructure capabilities that would otherwise require custom middleware :

• Caching — Identical requests are served from Cloudflare's global edge cache instead of hitting the xAI API repeatedly. This reduces latency and eliminates redundant token costs. Cache TTL is configurable per endpoint .
• Rate limiting — Teams can cap request volumes by rate, time window, and limiting method to prevent runaway bills and block suspicious traffic patterns .
• Analytics and logging — The dashboard surfaces request volume, token consumption, costs, and error patterns. Persistent logs are available on all plans, though storage limits vary: 100,000 logs total on the free tier, 10 million per gateway on paid plans .
• Fallback routing — When xAI's API returns errors or hits rate limits, AI Gateway can dynamically route requests to alternative models or providers without application-level changes .
• Unified billing — Multiple AI providers (OpenAI, Anthropic, now xAI, and others) are consolidated into a single Cloudflare invoice. Credits are purchased in the dashboard and drawn down across providers. This feature is currently in open beta .
• Secure key storage — API keys are held by Cloudflare's infrastructure rather than embedded in application code, reducing exposure risk .
• Guardrails — Content safety filters apply automatically alongside other gateway features, scanning prompts and completions .

Why this partnership matters strategically

The integration carries weight beyond a simple API connector announcement. It reflects how the AI infrastructure layer is consolidating, and where both companies are placing their bets.

For xAI, the partnership with Cloudflare provides enterprise-grade distribution without requiring xAI to build a global edge network itself. Cloudflare's infrastructure spans hundreds of cities, and routing Grok through that fabric means xAI's models inherit low-latency availability, caching, and observability that large enterprise customers expect. The move positions Grok as a multi-modal AI platform — text, code, image, video, and audio — that can compete directly with OpenAI and Anthropic for developer budgets .

For Cloudflare, adding the full Grok lineup strengthens the pitch that AI Gateway is a neutral, universal control plane. Every additional model family that plugs into the gateway makes Cloudflare's infrastructure stickier: once a team routes its AI traffic through Cloudflare for caching, logging, billing, and failover, switching costs rise. This directly competes with Vercel's AI Gateway and other middleware layers that aim to abstract model providers behind a unified API .

The timing is also notable. This announcement landed roughly one month after xAI released Grok 4.3 with its 1M-token context window and aggressive pricing, and just as the model reached general availability across multiple partner platforms. By locking in deeper distribution through Cloudflare at that moment, xAI gave developers a low-friction path to adopt the model alongside existing providers already routed through the same gateway .