Grok 4.3 Cuts Pricing While Gaining on Bench…

xAI released Grok 4.3 with input pricing roughly 40% lower and output pricing 60% lower than Grok 4.20, while scoring 53 on the Artificial Analysis Intelligence Index – four points above its predecessor. The biggest jump was on agentic tasks, where it hit 1500 ELO on GDPval-AA, surpassing Gemini 3.1 Pro Preview and Kimi K2.5. It uses 44% more output tokens than Grok 4.20 but still runs leaner than competing frontier models. The pricing move is aggressive: xAI is clearly trying to buy market share by undercutting on cost while incrementally improving capability. OpenAI, Google, and Anthropic have all compressed pricing on mid-tier models this year, but xAI is now applying the same pressure at the frontier tier.