Full Endpoint Accuracy Index results: GLM-5.2 provider benchmarks show 740% speed gap, 5.2x price gap

ArtificialAnlys · x · 2026-08-05

Artificial Analysis released full results for the Endpoint Accuracy Index, including detailed data for GLM-5.2, gpt-oss-120b, and DeepSeek V4 Pro. For GLM-5.2 (max), among 15 API providers, Databricks leads in output speed (388.4 t/s) and latency (6.18s), while DeepInfra (FP4) and CoreWeave offer lowest price ($0.49 per 1M tokens). Speed varies 740% and price up to 5.2x across providers.

Related event: Artificial Analysis Launches Endpoint Accuracy Index(4 posts)→

Original post →

More from Infra

Infra channel →