SpaceX unveiled Grok 4.5 on Wednesday, the first artificial‑intelligence model the company has trained specifically for software coding and autonomous agents. The release follows the $60 billion acquisition of the AI‑coding startup Cursor, whose data and editor were integrated into the training pipeline.
Musk’s team markets the model on cost efficiency: it consumes roughly half the tokens of comparable systems and is priced at $2 per million input tokens and $6 per million output tokens. Independent benchmark firm Artificial Analysis placed Grok 4.5 fourth on its GDPval‑AA v2 index with an Elo of 1543, noting it is “roughly comparable to Opus 4.7, but much faster.” The firm also measured a per‑task cost of $0.49, about 90 % cheaper than the leading rivals, putting the model on the Pareto frontier of performance versus price.
If the token‑efficiency holds up in real‑world codebases, enterprises may favor Grok 4.5 over Anthropic’s Claude or OpenAI’s flagship models, pressuring incumbents on high‑margin API traffic. The launch also showcases Musk’s vertically integrated AI stack—Cursor’s developer data, SpaceX’s Colossus supercomputer, and internal demand from Tesla and SpaceX engineering—raising both competitive and regulatory questions about the future of enterprise AI.
