xAI's Grok 4.5 launches under SpaceXAI with 1.5T parameters
Grok 4.5 achieves 83.3% on Terminal-Bench while using fewer tokens than Opus 4.8
xAI unveiled Grok 4.5 on July 8, 2026, a 1.5-trillion-parameter Mixture-of-Experts (MoE) model co-trained with Cursor coding data. The model achieves 83.3% on Terminal-Bench 2.1—a benchmark for terminal and command-line tasks—while using significantly fewer output tokens than Anthropic’s Opus 4.8. This efficiency improvement makes Grok 4.5 particularly attractive for cost-sensitive coding workflows. Pricing is set at $2 per million input tokens and $6 per million output tokens, undercutting many competitors. The release also marks the formal end of xAI as a separate entity, with all operations now consolidated under the SpaceXAI brand.
This unification signals a strategic shift for Elon Musk’s AI efforts, integrating xAI’s research into SpaceX’s broader technology stack. Grok 4.5’s co-training with Cursor data suggests a focus on developer tools, potentially challenging GitHub Copilot and other code assistants. The model’s high score on Terminal-Bench indicates strong performance in real-world terminal commands, scripting, and system administration tasks. For professionals, this means a specialized, cost-effective AI for coding and automation, backed by the resources of the SpaceXAI ecosystem. The dissolution of xAI as a standalone brand also simplifies the AI landscape, concentrating advanced models under one corporate umbrella.
- Grok 4.5 is a 1.5-trillion-parameter MoE model co-trained with Cursor coding data.
- Achieves 83.3% on Terminal-Bench 2.1 while using fewer output tokens than Anthropic's Opus 4.8.
- Priced at $2/M input and $6/M output tokens; marks xAI's dissolution into the unified SpaceXAI brand.
Why It Matters
Delivers a cheaper, more efficient coding model under the unified SpaceXAI ecosystem, targeting developers and automation workflows.