Elon Musk bought a coding startup for $60 billion and, in a move nobody could have possibly predicted, immediately shipped a model. Grok 4.5 is here, it's trained specifically for coding and agents, and it wants a permanent spot in your IDE.

A Frontier Model Built From an Acquisition

Grok 4.5 is the first joint release from SpaceXAI and Cursor, the AI coding platform SpaceX is in the process of acquiring. It runs on a 1.5-trillion-parameter mixture-of-experts foundation model, trained on tens of thousands of Nvidia GB300 GPUs, and is priced at $2 per million input tokens and $6 per million output tokens.

On benchmarks, it's putting up real numbers: 83.3% on Terminal-Bench 2.1, 64.7% on SWE-Bench Pro, and the top spot on Harvey's Legal Agent Benchmark. It's available now in Cursor, Grok Build, and the xAI API — though EU users are stuck waiting until mid-July.

The Real Fight Is Over Tokens, Not IQ

The headline number that should actually turn heads: Grok 4.5 reportedly uses roughly 4.2 times fewer output tokens than Anthropic's Opus 4.8 to complete the same SWE-Bench Pro tasks. In a world where API bills scale with tokens, efficiency is quietly becoming as important as raw capability.

This is also SpaceXAI's real coming-out party as a coding-model contender, arriving in the same week OpenAI dropped its GPT-5.6 lineup. The frontier model race isn't just about who's smartest anymore — it's about who's cheapest per solved bug.

Buying your coding tool's parent company is certainly one way to guarantee your model gets a warm reception in the editor.

Source: CryptoBriefing