Summary
Alex Finn shares his impressions of Claude Opus 5.5 after several days of early access from Anthropic. He argues it is the best model available today, ahead of Claude Fable and OpenAI’s ChatGPT 6 Astra, while also being faster and cheaper. He points to a reduction in list price on top of lower cost per task, and to benchmark charts in which Opus 5.5 on its high setting outperforms Astra on max.
Much of the video focuses on how the model communicates. Finn says that since roughly Opus 4.8, models from every major lab have drifted into odd jargon, with phrases like “smoke test” appearing constantly. In his experience Opus 5.5 brings back the concise, warm and readable style that made earlier Anthropic models pleasant to use, with clear tables and bullet points.
He also demonstrates its coding strengths using a complex game he has been building for a week. Given code originally written by Astra, Opus 5.5 flagged problems he had not asked about and offered to fix them. The video includes a sponsored segment from HubSpot for Startups on five installable Claude Skills for founder content, plus a look at how to get the most out of the new model.
📺 Source: Alex Finn · Published September 23, 2026
🏷️ Format: Review
