Epoch Benchmarks: GPT-5.6 Latency Grows Quadratically, Claude Near-Linear at Long Context

Epoch AI measured first-token latency up to 1M tokens: GPT-5.6's latency grows quadratically with context while Claude models remain near-linear, a key consideration for long-running agents.

2026-09-16 ~ 2026-09-17 · 2 related posts

• Epoch AI: GPT latency curves bend at long context while Claude stays linear — krishnan · 2026-09-16
• Epoch AI: GPT-5.6 latency scales quadratically with context, Claude 5 stays near-linear — dl_weekly · 2026-09-17