xAI launches Grok 4.5, undercutting Anthropic on coding-agent price

xAI released Grok 4.5 on July 8, 2026, pricing the coding-focused model at $2 per million input tokens and $6 per million output tokens, the company said.

That undercuts Anthropic’s Opus 4.8, which runs $5 and $25, though it does not beat OpenAI’s cheaper GPT-5.6 Luna at $1 and $6. It is xAI’s first model since SpaceX’s $60 billion acquisition of Cursor, and the first the company built for more than software engineering. The launch pushes the coding-agent price war into a fresh round.

Elon Musk, xAI’s founder, said Grok 4.5 is “an Opus-class model, but faster, more token-efficient and lower cost.” On xAI’s own benchmarks the model resolves tasks using an average of 15,954 output tokens, against a maximum of 67,020 for Opus 4.8 — a 4.2x efficiency gap.

Those figures are xAI’s own and are not independently verified. Its benchmark claims are mixed: the model beats Opus 4.8 on DeepSWE 1.0 and Terminal-Bench 2.1 but trails it on DeepSWE 1.1 and SWE-Bench Pro. Cursor, which co-announced the model and said it was trained on trillions of tokens of its own data, also disclosed that some of its codebase was accidentally included in training and has since been removed from future runs.

The model is live in Cursor, Grok Build and the SpaceXAI console, with a wider public release set for July 9. It is not yet available in the European Union.