Live Testing Sonnet 5.5 Vs Opus 5.5

Summary
Indian developer and streamer Rounit ("Rounieee") conducts an uncut multi-hour live stream testing Anthropic's newly released Claude Sonnet 5.5 against Claude Opus 5.5. Throughout the broadcast, he experiments with Sonnet 5.5 via the Claude Code CLI and Claude desktop/web apps, evaluating its capabilities on 3D Blender asset generation, WebGL rendering, and programmatic 2D canvas animation. He also reviews community benchmarks, API pricing differences, and viewer-submitted AI projects while interacting with live chat.

---

What is shown

• [01:20] Claude Code CLI updated and switched to test newly deployed models, including Opus 5.5 and Sonnet 5.5.
• [03:15] Reviewing benchmark charts on X comparing Sonnet 5.5, Sonnet 5, Opus 5.5, and GPT-6 Sol on Agentic Coding (Terminal Bench 4.0, CursorBench 4.0, FrontierCode 1.1) and GPQA.
• [12:35] Prompting Claude Sonnet 5.5 in the desktop app to build a realistic 3D Formula 1 racing car using modular Blender Python scripts and Blender MCP.
• [34:50] Examining official Anthropic pricing and speed metrics comparing Sonnet 5.5 ($2/$10 per million tokens) to Opus 5.5 ($4/$20).
• [36:00] Testing a Three.js interactive 3D sunset beach simulation generated by Sonnet 5.5 in the browser at localhost:5173.
• [56:30] Prompting Sonnet 5.5 in the CLI to generate a pure JavaScript/HTML5 Canvas hand-drawn animation tracing the evolution of humans from early hominids to modern humans.
• [66:55] Displaying the initial prototype of the human evolution animation running locally at localhost:8137.
• [82:00] Inspecting the raw 3D mesh wireframes generated inside Blender 3.6 for the Formula 1 car project.
• [95:05] Reviewing the rendered 3D F1 race car viewer in-browser (the "Kestrel 77" with Voltrix livery, interactive angles, auto-rotation, wheel rolling, and customizable color schemes).
• [125:20] Viewing a community showcase web project (hawkapp.in) featuring interactive paper-style animation built using Claude.
• [158:05] Running the refined version of the "Human Evolution in Under One Minute" canvas animation, featuring colored scenes, walking animations, era transitions (7 million years ago to modern day), and background environments.

---

Claims & numbers

• Sonnet 5.5 Benchmarks: According to benchmark tables displayed on X, Claude Sonnet 5.5 scores 70.6% on Terminal Bench 4.0 (compared to Opus 5.5's 66.4% and Sonnet 5's 10.3%), 46.2% on FrontierCode 1.1, and 58.4% on CursorBench 4.0.
• Speed and Pricing: Anthropic's comparison chart states that Sonnet 5.5 is 30%+ faster than Sonnet 5 and priced at $2.00 per million input tokens and $10.00 per million output tokens ($0.20 cache read, $2.50 cache write), exactly half the cost of Claude Opus 5.5 ($4.00 input / $20.00 output).
• Token Consumption: The presenter notes the 3D F1 car build used roughly 44,000 tokens over ~35–40 minutes of iterative coding, while the multi-stage human evolution animation consumed over 70,000 tokens during development.
• Personal Business Metrics: Rounit states he scaled his SaaS project (a library management platform) to $8,500 MRR with 250–300 active subscribers over 8.5 months, and utilizes a $20/month Claude Pro plan.

---

Notable quotes

• [00:13] "We are going to test Sonnet 5.5, the latest model that released today... right now... directly with Opus 5.5."
• [34:52] "Sonnet 5.5 requires fewer tokens per task than Sonnet 5, so it's less expensive to run. It also generates output 30%+ faster..."
• [95:13] "Sonnet 5.5 in one shot, crazy 3D model, not gonna lie."

---

Assessment
This is an authentic, unedited live stream demonstrating hands-on developer testing of Claude Sonnet 5.5 immediately following its launch. The coding runs, terminal outputs, browser renders, and session token usage are shown live with real-time trial, error, and verification.

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