Atlas

Benchmarks

← All benchmarks

BridgeBench V3 Arena Speed Throughput

Indexes · 2026-07-10

The BridgeBench V3 Speed arena reports median output throughput while paired models answer the same tasks concurrently. Higher token throughput is better, and results are measured directly without model judges.

Top models (higher is better)

ModelScore
Grok 4.5674
GLM-5.2340
Claude Fable 5106
GPT-5.6 Sol90.1
Loading Atlas data…