Grok 4.5 launches at $2/$6 per million tokens, undercutting Opus 4.8 on cost and speed
SpaceXAI and Cursor released Grok 4.5 on July 8–9, 2026, the first AI model trained jointly by Elon Musk's AI unit and Cursor following SpaceX's $60 billion acquisition of the coding startup. Priced at $2 per million input tokens and $6 per million output tokens, with a faster variant at $4/$18, the model achieves roughly 2x the token efficiency of comparable leading models.
On SWE-Bench Pro tasks, Grok 4.5 needs around 15,900 output tokens on average—roughly 4.2 times fewer than Opus 4.8. On Terminal Bench 2.1, Grok 4.5 scores 83.3%, trailing only Fable (84.3%) and GPT-5.5 (83.4%), ahead of Opus 4.8 (78.9%).
The model is live in Grok Build, across all Cursor plans (desktop, web, iOS, CLI, SDK), and via the SpaceXAI console, with Cursor offering double usage for the first week. Grok 4.5 is not yet available in the EU, with SpaceXAI expecting access to open in mid-July.
For teams running cost-sensitive inference pipelines, the ~60% price discount vs. Opus 4.8 and token efficiency claim warrants a test—though builders chasing top absolute performance should still prioritize Fable 5 and Opus 4.8 on the hardest benchmarks.
Sources
- Primary source
- Introducing Grok 4.5 | SpaceXAI
“$2 per million input tokens and $6 per million output tokens; 2x token efficiency”
- SpaceXAI And Cursor Release Grok 4.5
“First model built jointly with Cursor; 4.2x token efficiency vs. Opus 4.8 on SWE-Bench Pro”