SPACEXAI DEBUTS GROK 4.6, OVERTAKING KIMI K3'S PERFORMANCE AND MATCHING GPT-5.6 SOL FOR WORLD'S THIRD BEST ON ARTIFICIAL ANALYSIS

Summary from the TLDR newsletter

SpaceXAI has released Grok 4.6, a model focused on long-running agents, coding, and knowledge work. The model surpasses Kimi K3 and ties with GPT-5.6 Sol Max on the Artificial Analysis Intelligence Index. It posts sizable gains over its predecessor across coding, terminal, knowledge-work, and agent benchmarks. The API starts at $2 per million input tokens and $6 per million output tokens, less than half of what GPT-5.6 Sol costs in standard mode.