Grok

Overview

This is a model family overview. For version-specific details, see the individual model entries linked below.

Grok is xAI’s model family for reasoning-heavy assistants, tool-using agents, coding routes, and the consumer Grok product. Grok 4.5 is the July 2026 flagship for coding, agentic tasks, knowledge work, and technical reasoning. Grok 4.3 remains a lower-cost general route, while Grok Build 0.1 remains a separately priced coding API model.

SpaceX acquired xAI in February 2026, and current launch pages use SpaceXAI as public masthead and product branding. The xAI API domain, model identifiers, environment variables, legal terms, and footer identity still use xAI or X.AI LLC, so both names remain relevant: SpaceXAI for the combined-company brand and xAI for the technical and legal identity.

Current Latest

Grok 4.5 is the current flagship and the default model in Grok Build. It accepts text and images, supports tool use and structured outputs, exposes a 500K-token context window, and is priced at $2 per million input tokens and $6 per million output tokens. xAI also documents a $0.50 cached-input rate and higher pricing above 200K tokens.

Grok 4.3 remains a current, lower-cost general route with a 1M-token context window and $1.25 input / $2.50 output pricing. Grok Build 0.1 remains a 256K public-beta coding API model at $1 input / $2 output, but it is no longer the Grok Build CLI’s default model.

The most relevant Grok entries at this snapshot:

• Grok 4.5 as the flagship for coding, agentic knowledge work, science, engineering, and tool-using reasoning.
• Grok 4.3 as a lower-cost 1M-context route for general agentic and assistant work.
• Grok Build 0.1 as a cheaper pinned coding API route and migration target for old Grok Code Fast configurations.
• Grok 4.20 as a 1M-context preview/enterprise research lane for teams explicitly testing the beta and multi-agent family.
• Grok Imagine for image and video generation through xAI’s media endpoints.
• Grok 4 and Grok 4.1 Fast as deprecated snapshots whose old slugs now redirect to Grok 4.3 with documented reasoning-effort defaults.
• Grok Code Fast 1 as a deprecated coding snapshot whose explicit replacement is Grok Build 0.1.

Strengths

The family is strongest when a team wants xAI’s combination of reasoning, tool calling, and live-information access. xAI’s current docs also make it clear that Grok is designed to work with web search, X search, code execution, collections, connectors, and MCP-style tool flows rather than remaining a plain text model.

Pricing segmentation is also clearer than before. Grok 4.5 is the higher-capability flagship, Grok 4.3 is the lower-cost general route, and Grok Build 0.1 is the cheaper coding API route. Requests to deprecated text slugs can still redirect to newer models, so explicit model selection is safer than leaving old aliases in production.

When to Choose Grok

Choose Grok when the workflow depends on tool calls, current information, or xAI-native product alignment. Start with Grok 4.5 for difficult coding and agentic knowledge work. Evaluate Grok 4.3 when price or its larger context window matters more, and use Grok Build 0.1 when a cheaper pinned coding API route is the better fit.

For teams that need slower-moving documentation, stronger governance defaults, or less product churn, Anthropic and OpenAI remain easier operational baselines. Grok is improving quickly, but it still benefits from closer monitoring than the most mature alternatives.

Access

• xAI API
• Grok web and mobile apps
• X ecosystem consumer surfaces
• xAI enterprise and beta access paths for newer Grok releases
• partner surfaces and gateways where individual models are enabled

EU availability can lag global launches. As of July 10, xAI said Grok 4.5 had not yet reached its EU product and API-console surfaces, so teams should verify account-level access before planning a rollout.