grok-3

Grok 3 is xAI's flagship model, supporting a 131K-token context window, vision inputs, tool use, and streaming. A key differentiator is real-time web access (available via API with the search tool) — enabling it to answer questions about current events without the knowledge cutoff limitations of standard models. It is available via xAI's API.

Grok 3 is priced at $3.00 per million input tokens and $15.00 per million output tokens. At 1,000 input + 500 output tokens per call, it costs approximately $0.0105 per 1,000 API calls, placing it in the premium advanced tier. For applications requiring real-time web access, the cost premium over models without this capability may be justified — though web search adds latency and the search tool must be explicitly enabled.

Grok 3's 131K-token context window handles most production use cases. Median latency is approximately 400–700ms TTFT under typical load. Throughput is approximately 80 tokens per second. Maximum output is 16,384 tokens. Real-time web search adds 1–3 seconds of latency when enabled, as the model must retrieve and process search results before generating a response.

Grok 3 is best suited to applications requiring real-time information (news, current events, live data), general-purpose chat and reasoning, coding assistance, and workloads where X/Twitter integration is valuable. Its real-time web access is its primary differentiator — for applications where knowledge cutoff is a significant limitation, Grok 3 with search enabled is one of the few production options. Vision support handles images alongside text. Tool use is available for agentic workflows. For applications that don't require real-time information, other advanced-tier models offer comparable capability with more mature API ecosystems.

Grok 3 sits at the top of xAI's model lineup, above Grok 3 mini (a smaller, faster variant). Its main competitors are advanced-tier proprietary models from OpenAI and Anthropic, and other web-connected models. xAI's API is newer and has a smaller ecosystem than established providers, which may be a consideration for teams evaluating production reliability and tooling support. Its intelligence score of approximately 83/100 reflects strong advanced-tier performance, with the real-time web access capability adding unique value for specific use cases.