Breaking

Developer Test Shows Claude Opus 5.5 Outperforming Sonnet 5.5 on a Blender Whale Build at Higher Cost

September 29, 2026 at 21:46 EDT

Foundation Models
AI Agents
Media Generation

A public developer comparison has put Anthropic's newest Claude models into a practical 3D-creation test: build a steampunk whale in Blender, then move it into three.js for browser playback. Claude Opus 5.5 was judged the stronger result, but the run cost about $156 in API-equivalent usage versus about $109 for Claude Sonnet 5.5.

Sep. 30, 2026 · Claude 5.5 comparison

Opus still wins the whale-off

A structured Blender build gave Claude Opus 5.5 the stronger steampunk whale, while Claude Sonnet 5.5 came in cheaper and still impressed developers as a practical creative workhorse.

Reported run cost, drawn to scale

~$156Claude Opus 5.5

~$109Claude Sonnet 5.5

Each block represents roughly $10 of API-equivalent cost.

2.50MOpus output tokens

1.68MSonnet output tokens

1MContext window for both

128KMax output for both

Workflow, not one-shot magic

Structured promptsSame prompt series

→

Blender build3D scene creation

→

three.js exportBrowser playback

What impressed developers

Opus was seen as more detailed and better engineered, while Sonnet looked strong enough to be useful for lower-cost iteration.

What limited the result

The build relied on tool setup, agent workflow and repeated prompting, so it does not prove a simple prompt can reproduce the same quality.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.