Grok 4.3

Grok 4.3 is a proprietary large language model from xAI positioned as a step beyond its direct predecessor, Grok 4.2, with notable gains on third-party evaluations reported by Artificial Analysis. The model is purpose-built for enterprise-grade AI agents that must reason over very long inputs, use tools reliably in customer-facing scenarios, and parse complex documents such as legal and financial materials. xAI markets the release as a flagship advancement aimed at keeping pace with frontier offerings from OpenAI, Anthropic, Google, and major Chinese labs.

According to xAI, Grok 4.3 tops the Artificial Analysis Omniscience benchmark with the lowest hallucination rate among frontier models, leads Artificial Analysis Tau2 Telecom for real-world tool calling in customer support agents, and ranks first on the Vals AI Case Law and Corporate Finance benchmarks for complex document understanding. Developers can dial the model's reasoning effort across none, low, medium, and high settings, trading latency for depth on harder prompts. The release is framed as a strong fit for organizations building secure, agentic applications where long-context retrieval, accurate tool use, and trustworthy document reasoning matter more than raw conversational flair.