Unverified50% confidenceBenchmarkExact time
据Artificial Analysis测试，Grok 4.6在MXX High档位约消耗38k tokens每任务，而Grok 4.7飙升至81k tokens
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/25/2026
First Seen
Valid until: 12/24/2026
Sources
Related Entities
Related Claims
Unverified在Gemma 4 12B模型的Token生成测试中，Spark达26.4 tokens/s、Halo 24.6 tokens/s、M4 Pro最快33.8 tokens/s76% similarUnverified根据Artificial Analysis数据，Fable完成单个任务约用33k token，Opus 5约用37k token75% similarUnverified据称 Grok 4.5 的 token 效率约为同类领先模型的两倍，能用不到一半的步骤完成任务73% similarUnverifiedGrok 4.5在SWE Bench Pro任务上的输出Token比Opus 4.8少4.2倍72% similarUnverified投机采样中当候选 token 的接受率较高时，每次前向传播的有效输出 token 数量可达 2~4 个，从而将整体生成速度提升 1.5x~3x72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/948934API
curl https://kongchang.com/api/v1/knowledge/claims/948934MCP
get_claim(id=948934)