Anthropic Just Dropped Claude Opus 5.5 (CHEAPER & BETTER)

Summary
Brock Mesarich reviews Anthropic's release of Claude Opus 5.5, breaking down its cost reductions, performance benchmarks, and speed improvements. He highlights Anthropic's benchmark comparisons against models like Claude Fable 5.1 and GPT-6 Astra, and tests Opus 5.5's new communication style against his own YouTube channel analytics.

What is shown

• [00:00] Screen recording of Anthropic's announcement website and an "AI Weekly" summary newsletter for Claude Opus 5.5.
• [00:24] Breakdown of running costs and API pricing tables ($4/M input, $20/M output, $0.20/M cache reads).
• [01:22] Anthropic benchmark comparison chart showing scores across Agentic coding, GDPval-AA v1.1, OSWorld 2.0, ChartQA Pro, and more against Claude Fable 5.1, Claude Opus 5, GPT-6 Astra, and GPT-5.6 Sol.
• [02:31] Case study graphics: C-to-Rust HAProxy migration (9.5 hours vs. 12 hours) and financial spreadsheet plus executive presentation generation (63 minutes vs. 93 minutes).
• [03:37] Side-by-side text comparisons between Claude Opus 5 and Opus 5.5 on bug explanations, Slack thread summarization, and design change explanations.
• [05:39] Hands-on test by the presenter comparing Fable 5.1 and Opus 5.5 parsing his YouTube metrics on an Excalidraw board.
• [06:22] Demonstrations of user creations shared by Anthropic: an animated watermelon short story, an interactive Apollo 8 Earthrise simulation, and an interactive playable catapult pencil sketch.
• [07:14] Usage limit updates (higher 5-hour limit and saveable rate limit resets) and deployment platforms (Claude, Claude Code, Claude Platform, AWS, GCP, Azure).

Claims & numbers

• Anthropic released Claude Opus 5.5 on September 22, 2026, as the first model of the Claude 5.5 family (the presenter notes).
• Opus 5.5 runs at 40% lower operational cost compared to Opus 5 while matching or exceeding Claude Fable 5.1 performance (the presenter states).
• API pricing: Input tokens reduced from $5 to $4 per million; output tokens reduced from $25 to $20 per million; cached input reads reduced from $0.50 to $0.20 per million; cache writes are $5 per million (the presenter shows).
• Output generation is reported to be over 30% faster than Opus 5 while requiring less compute to serve (the presenter notes).
• Benchmarks shown include:
  • Agentic coding (Terminal-Bench 4.0): Opus 5.5 at 66.4% vs. Fable 5.1 at 55.8%, Opus 5 at 52.3%, GPT-6 Astra at 57.3%, and GPT-5.6 Sol at 37.3%.
  • FrontierCode v1.1 (Main): Opus 5.5 at 54.4% vs. Fable 5.1 at 50.3%.
  • CursorBench 4.0: Opus 5.5 at 57.6% vs. Fable 5.1 at 51.8%.
  • Knowledge work (GDPval-AA v1.1): Opus 5.5 at 1846 vs. Fable 5.1 at 1725, GPT-6 Astra at 1542.
  • Computer use (OSWorld 2.0): Opus 5.5 at 81.5% vs. Fable 5.1 at 80.7% and Opus 5 at 74.0%.
  • Visual chart recognition (ChartQA Pro): Opus 5.5 at 89.0% vs. Fable 5.1 at 88.4%.
  • GPT-6 Astra leads Opus 5.5 in Business workflows (AutomationBench: 41.4% vs. 40.0%) and Scientific research (Terminal-Bench Science 0.1: 64.4% vs. 58.7%).
• In an internal test migrating HAProxy C to Rust, Opus 5.5 took 9.5 hours with a reported 51% cost reduction compared to Fable 5.1's 12 hours (the presenter shows).
• In a spreadsheet and presentation task, Opus 5.5 finished in 63 minutes at 50% lower cost compared to Opus 5's 93 minutes (the presenter shows).
• Anthropic increased 5-hour usage limits across Pro, Max, Team, and seat-based Enterprise tiers and added a saveable rate limit reset (the presenter states).

Notable quotes

• [00:26] "First things first, we have 40% lower costs with this model compared to the previous Opus 5 model."
• [02:50] "It's not necessarily, in my opinion, all about the new capabilities that it unlocks, rather how cheap can you run specific tasks compared to other models."
• [03:55] "Its messages are much easier to understand at a glance, which testers said helped during long working sessions."

Assessment
This is an independent creator review and walkthrough reacting to Anthropic's official blog post and launch materials for Claude Opus 5.5. Most data points are directly cited from Anthropic's published announcements and graphics, supplemented by a simple real-world text formatting comparison performed by the creator on his own channel data.

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