Elon Musk's xAI released Grok 4.5, positioning the 1.5 trillion parameter model as "Opus-class" but faster and more cost-efficient than Anthropic's flagship.

The model represents xAI's first entry into frontier coding and agent workflows, trained in partnership with Cursor. The AI code editor called it "our most powerful model yet" and "the first we've built for more than software engineering."

Cursor announced immediate in-product availability with double usage for the first week. The company clarified that Grok 4.5 operates in a different weight class from its smaller Composer series models.

Grok 4.5 costs $2 per million input tokens and $6 per million output tokens, significantly undercutting competitors. For comparison, GPT-5.6 costs $5 input and $30 output, while Opus 4.8 runs $5 input and $25 output per million tokens.

The model features a 500,000 token context window, down from Grok 4.3's one million tokens. Musk said the context limit would likely return to one million tokens "by next week."

Performance benchmarks

Artificial Analysis ranked Grok 4.5 fourth on its Intelligence Index with a score of 54, behind only Fable 5, GPT-5.5, and Opus 4.8. The model gained 16 points over Grok 4.3 on the same benchmark.

On coding tasks, Grok 4.5 scored 76 on the Coding Agent Index in Grok Build, matching GPT-5.5 performance in Codex but trailing Fable 5 in Claude Code.

The model's efficiency advantage showed in token usage. Grok 4.5 averaged 14,000 output tokens per Intelligence Index task, 60% lower than Opus 4.8. On coding tasks, it used 1.9 million total tokens versus 7.2 million for Fable 5.

Early ecosystem support appeared immediately, with day-zero availability in Hermes Agent, OpenRouter, and existing Grok subscriptions. The launch comes one day before OpenAI's confirmed GPT-5.6 release.