Claude Opus is 10x faster than OpenAI GPT 5 at non-streaming completions

This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).

A developer compared non-streaming completion latency across Claude claude-opus-4.8, OpenAI gpt-5, and gpt-5-mini for a real-world blog comment rewriting task. Claude was nearly 10x faster than gpt-5, while gpt-5-mini was both faster and cheaper than gpt-5 (3x price difference). As a bonus, the author also measured a latency difference between using the litellm wrapper versus the native OpenAI SDK, and decided to drop litellm in favor of native SDKs partly due to recent CVEs.

2m read timeFrom peterbe.com
Post cover image
Table of contents
ConclusionBonus
268 Impressions