Grok 4.5: 80% Cheaper Coding Agent, Near-Frontier Speed

Grok 4.5 dramatically cuts the cost of agentic coding work. It undercuts Anthropic's Claude Opus 4.8 by over 60% on headline rates. For a completed agentic task, Grok 4.5 costs $2.49, compared to $5.07 for GPT-5.5 and $11.80 for Fable 5. This means significant savings for teams running many sessions. Grok 4.5 ranks fourth on Artificial Analysis's Intelligence Index. This is a 16-point jump from its predecessor, Grok 4.3, marking the largest single-generation leap recorded. It ties with GPT-5.5 on the Coding Agent Index. The model achieves its lower cost and speed due to its mixture-of-experts architecture. This design allows it to run at 80 tokens per second and resolve coding tasks with fewer output tokens than competing models. This efficiency is a direct result of its construction, not a temporary discount. This development changes the economics for businesses using AI for coding tasks, making advanced models more accessible.