Grok 4.6 hits 1,753 Elo on GDPVal-AA, claiming top non-Anthropic model spot

SpaceXAI's Grok 4.6 matches OpenAI's best model and undercuts it on price

Grok 4.6 scores 61 points (tied with GPT-5.6 Sol) while costing 60% less ($2/$6 per M tokens). Dominates agentic multi-step tasks: 53 steps vs Claude Opus 5's 103. For IT teams optimizing LLM spend, price-performance shifts economics toward agent automation; Grok 4.6 becomes the efficient choice unless peak reasoning (Claude Opus 5, 63pts) is essential.