Opus 5.5 Is Here - Claude Is So Back!

Summary — In this video, content creator Paul breaks down the release of Anthropic's Claude Opus 5.5, announced on September 22, 2026. He reviews Anthropic's announcement posts, pricing structure, effort settings in the web interface, benchmark performance against rival models, and changes to usage limits.

What is shown

• [00:04] Slide displaying the launch title "Claude Opus 5.5" dated September 22, 2026.
• [00:18] The Claude web application interface showing the model picker dropdown, featuring Fable 5.1, Opus 5.5, Sonnet 5, and Haiku 4.5.
• [00:26] Anthropic's post on X introducing Claude Opus 5.5 and detailing cost/performance comparisons against Fable 5.1 and Opus 5.
• [01:12] Pricing table comparing Claude Opus 5.5 against Claude Opus 5 per 1M tokens, along with AutomationBench charts.
• [01:50] The Claude model settings interface demonstrating that Opus 5.5 defaults to "Medium" effort while Opus 5 defaults to "High" effort.
• [02:52] A brief prompt submitted to Opus 5.5 asking "What can you tell me about the new opus 5.5?".
• [03:07] A comprehensive benchmark comparison table contrasting Claude Opus 5.5 (at max effort) against Fable 5.1, Opus 5, GPT-6 Astra, and GPT-5.6 Sol across coding, knowledge work, reasoning, and computer use.
• [04:25] A side-by-side text comparison of Claude Opus 5 versus Opus 5.5 explaining a billing bug to show differences in conversational tone.
• [05:03] Anthropic's announcement tweet regarding increased five-hour rate limits and a banked rate limit reset feature for Pro, Max, and Team plans.

Claims & numbers

• Release and Availability: The presenter states Claude Opus 5.5 was released on September 22, 2026, across the API, web app, and desktop app.
• Cost and Speed: Anthropic claims Opus 5.5 performs at the level of Claude Fable 5.1 for most tasks, costs 40% less to run at default effort settings than Opus 5, and generates outputs more than 30% faster than Opus 5.
• Pricing per 1M tokens (Claude Opus 5.5 vs Opus 5):
  • Input tokens: $4 (vs $5 for Opus 5)
  • Output tokens: $20 (vs $25 for Opus 5)
  • Cache reads: $0.20 (vs $0.50 for Opus 5)
  • Cache writes: $5 (vs $6.25 for Opus 5)
• Effort Setting Distinction: The presenter highlights that Opus 5.5 defaults to "Medium" effort, whereas Opus 5 defaults to "High" effort, which affects cost and speed metrics.
• Benchmarks (Opus 5.5 at max effort):
  • Terminal-Bench 4.0 (Agentic coding): 66.4% (vs Fable 5.1 at 55.8%, Opus 5 at 52.3%, GPT-6 Astra at 57.9%, GPT-5.6 Sol at 37.3%).
  • FrontierCode v1.1: 54.4% (vs Fable 5.1 at 50.3%, GPT-6 Astra at 53.3%).
  • CursorBench 4.0: 57.8% (vs Fable 5.1 at 51.8%).
  • GDPval-AA v2.1 (Knowledge work): 1846 (vs Fable 5.1 at 1735, Opus 5 at 1708, GPT-6 Astra at 1542).
  • AutomationBench (Business workflows): 40.0% (vs Fable 5.1 at 31.4%, Opus 5 at 26.9%, GPT-6 Astra at 41.4%).
  • Humanity's Last Exam (Reasoning): 67.7% with tools (vs Fable 5.1 at 65.6%, Opus 5 at 63.6%, GPT-6 Astra at 57.2%).
  • Terminal-Bench-Science 0.1: 58.7% with tools (vs Fable 5.1 at 52.6%, GPT-6 Astra at 64.6%).
  • OSWorld 2.0 (Computer use): 81.8% partial (vs Fable 5.1 at 80.7%, Opus 5 at 74.0%).
  • Chartography (Visual chart recognition): 89.0% (vs Fable 5.1 at 88.4%, Opus 5 at 83.4%).
• Usage Limits: Anthropic announced an increase to five-hour usage limits on Pro, Max, and Team subscriptions, alongside a saveable banked rate limit reset.

Notable quotes

• [00:00] "Claude Opus 5.5 is here. And I was not expecting this, but from the looks of it, Claude is back."
• [01:31] "But there's something a little bit off here, because for both of these claims, it says 'at its default effort settings.'"
• [04:41] "Technical language—it is very typical AI response language. Over here though, if you look at 5.5, it feels a lot more natural."

Assessment This video is a third-party commentary and overview of Anthropic's official announcement and documentation. While the presenter demonstrates the model selection UI and shows official benchmark tables, he does not perform live benchmark replications or extensive hands-on testing during the video, relying primarily on Anthropic's published materials.

Described by gemini-3.8-flash on 2026-09-29 from the video's audio and frames.