Grok 4.5 claims 4.2x token efficiency over Claude Opus 4.8, verified by third party

SpaceXAI's new Grok 4.5 model claims significantly higher token efficiency compared to Anthropic's Claude Opus 4.8. While SpaceXAI reported Grok 4.5 uses 4.2x fewer output tokens on the SWE Bench Pro, independent analysis from Artificial Analysis corroborates this, finding Grok 4.5 uses over 60% fewer tokens on its Intelligence Index. This efficiency, combined with lower per-token pricing, suggests Grok 4.5 could be substantially cheaper for certain tasks. AI

IMPACT Potentially lowers operational costs for AI tasks, challenging competitors on efficiency metrics.

RANK_REASON New model release from a frontier lab (SpaceXAI) with specific performance claims. [lever_c_demoted from frontier_release: ic=1 ai=1.0]

Read on Towards AI →

• Anthropic
• Artificial Analysis
• Claude Opus 4.8
• Grok 4.5
• SpaceXAI
• SWE Bench Pro

• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →