Grok 4.6 is 'objectively No. 1' — and 85% cheaper than rivals, Musk claims

SpaceX has unveiled Grok 4.6, a new language model that Elon Musk claims is "objectively No. 1 in terms of intelligence, speed, and cost."

Investor Gavin Baker wrote on X that Grok 4.6 delivers roughly the same performance as Fable 5 Max but at an 85% discount. Input tokens are 80% cheaper, and output tokens are 88% cheaper. Baker added that Grok 4.7 will be a significant step up, as it is a much larger model incorporating Cursor and SpaceX data for pretraining.

Independent analytics firm Artificial Analysis confirmed the leap, reporting that Grok 4.6 scores 61 points on its AI index, approaching the level of GPT-5.6 Sol, while demonstrating outstanding agent performance at a lower cost.

Pricing for Grok 4.6 is set at $2 per million input tokens and $6 per million output tokens — more than 60% below Claude Opus 5 ($5/$25) and GPT-5.6 Sol ($5/$30). The cost per single task is $0.84.

The first xAI chatbot launched in November 2023 as a direct rival to ChatGPT, featuring real-time access to X data and a focus on humor and unfiltered answers. Since then, the model has come a long way. In July 2025, Grok 4 first put xAI at the top of Artificial Analysis' Intelligence Index, overtaking OpenAI, Google, and Anthropic. That same month, the multi-agent version Grok 4 Heavy posted record results on the Humanity's Last Exam benchmark.