elon musks xai has launched grok 4 fast a cheaper ai model handling reasoning and nonreasoning tasks together it uses 40 fewer tokens while keeping accuracy and supports a 2 million token context window scoring high on benchmarks it costs 020 per million input tokens it offers up to 98 savings over grok 4