The AI coding agent market is on fire. Five players are battling for the same space. And now SpaceX — Elon Musk’s rocket company — holds an option to buy Cursor for 60 billion dollars.
The story, reported by The Verge in April 2026, is a purchase option with a 10-billion-dollar penalty clause if SpaceX decides not to exercise it. An aerospace company is betting big on software that writes software.
Cursor just launched Composer 2.5, built on the Kimi K2.5 model. The company published its own numbers, and on the three main coding benchmarks it matches Claude Opus 4.7. Terminal-Bench 2.0: 69.3% versus 69.4%. SWE-Bench Multilingual: 79.8% versus 80.5%. These are statistical ties. The difference is in price: Cursor costs roughly 2 dollars per task and Claude Opus 4.7 costs 10, five times more.
OpenAI launched Codex CLI in February 2026 as a direct answer to Claude Code’s success. xAI entered in May with Grok Build, a terminal agent with parallel sub-agents. And Alibaba unveiled Qwen 3.7 Max, available on OpenRouter at $2.50 per million input tokens, dramatically cheaper than Western alternatives.
But not everything circulating on YouTube is true. A video from the AI Skills Hub channel claims Qwen “outperformed GPT and Gemini in coding tests.” There is no evidence backing that claim. In fact, GPT-5.5 scores 82.7% on Terminal-Bench 2.0, well above Composer 2.5 (69.3%) and Opus 4.7 (69.4%). The same channel asserts Grok was trained on Cursor interaction data. Zero evidence. It’s a fabrication.
Claude Code is in the lead. Cursor is closing the gap on price. SpaceX is putting its Colossus supercomputer — “one million H100 equivalents” — behind this bet. OpenAI counterattacks with Codex. Google watches with Gemini CLI. And the Chinese alternatives apply pressure on cost.
The direction is clear: coding agents are becoming a commodity. The edge now lies in who controls the ecosystem, the infrastructure, and the developer workflow.
Main source: Introducing Composer 2.5 — Cursor Blog