GROK 4.3
Released April 30, 2026. 1M-token context. Vision input. Native X / web / news search. Strong reasoning at $1.25/$2.50 per million tokens — a fraction of Grok 4's launch price.
Grok 4.3 is xAI's flagship as of late April 2026. It ships with a 1M-token context window, vision input, and the most differentiated web tool in the frontier set: native X (Twitter), web, and news search grounded against xAI's own crawler. Reasoning quality lands close to GPT-5.5 on hard benchmarks while pricing comes in at $1.25 per million input tokens and $2.50 per million output — roughly a quarter of GPT-5.5's price and an order of magnitude cheaper than Claude Opus 4.7. The case for Grok 4.3: real-time information from X and the open web (it sees posts the moment they're indexed), long-context work, and cost-sensitive deployments where you still want frontier-class reasoning. The case against: the tool-calling ecosystem is thinner than OpenAI's, refactor-heavy coding still favors Claude Sonnet 4.6 / Opus 4.7, and writing voice trails Opus. For most general work that needs current information, Grok 4.3 is the new value pick.
On most benchmarks Grok 4.3 matches or beats Grok 4 while costing a fraction as much. It's the obvious default if you were on Grok 4.
$1.25 per million input tokens and $2.50 per million output tokens. Council AI bundles it inside monthly plan budgets.
Yes — Council AI is fully managed and ships Grok 4.3 on every paid tier. No xAI account required.
Grok grounds against xAI's own crawler with first-party access to X (Twitter). For real-time chatter, breaking news, and what's-happening-now queries, no other frontier model has the same fresh signal.
Grok 4.3 is the daily-driver flagship. Pick Grok 4.20 only when you need the 2M-token window for huge codebases or very long documents.
Yes, with native tool calling. It's reliable for most agent flows, though GPT-5.5 still leads on the hardest structured-output cases.