Claude Opus 5 Review: Better Apps at Lower Cost

TL;DR

Claude Opus 5 produced functional interactive websites, games, simulations, editors, and data visualizations with little manual effort in the creator's tests. The creator found its outputs more polished and less expensive than Fable 5 for comparable website tasks, while reported benchmark results also placed it ahead of several referenced frontier models in agentic terminal coding and knowledge work.

Transcript

Anthropic just dropped Claude Opus 5 and it is by far probably the best Frontier large language model currently available to us underassmen. In this video, I'm going to show you everything that you need to know about how to take the most advantage and use Opus 5 for the best of its capabilities. This is a website that essentially Opus created itsel... Read More

Key Insights

• Claude Opus 5 can generate complete interactive 3D experiences with minimal manual work, according to the creator's testing. One example was a navigable virtual art gallery with generated artwork, mouse-over sounds, and a cataloging mechanic that made the site resemble a small game.
• Simulation development is a notable strength demonstrated by Opus 5. The tested applications included orbital mechanics, deformable cloth, cellular automata, predator-prey populations, quadcopter flight, pendulum motion, and destructible structures, often combining explanatory content with controls that let users experiment directly.
• Opus 5 can create application features beyond static visual presentation. The pixel editor included drawing, erasing, undo, onion-skin functionality, multiple layers, frames, animation capabilities, and adjustable canvas dimensions, suggesting that the model can assemble several related tools into one usable interface.
• Generated assets can reduce dependence on downloaded resources. In the showcased product website and sneaker configurator, Opus 5 created visual elements itself, including SVG resources, rather than simply retrieving existing images, although the sneaker representation did not always look clearly like a shoe.
• Model-generated simulations can appear convincing without guaranteeing scientific accuracy. The cloth demonstration responded plausibly to pulling and tearing, but the creator explicitly avoided claiming that its fabric dynamics were fully accurate, distinguishing visual credibility from validated physical correctness.
• Opus 5 still produces visible mistakes despite its strong overall performance. The creator identified an apparent design error in the double-pendulum interface, while resizing the pixel editor caused an unhappy-looking result and the sneaker configurator produced an object with an unclear shoe-like appearance.
• Opus 5 was cheaper than Fable 5 in the specific website comparison shown. The Opus-generated Airlines site reportedly cost 69 cents, compared with 94 cents for the Fable 5 version, and the creator judged the Opus output to be more sophisticated and visually compelling.
• Reported benchmark results placed Opus 5 strongly across coding, knowledge work, and novel problem solving. The transcript cites a GDPval-AA V2 score of 1861 and an ARC-AGI 3 score of 30.2%, compared with 1.5% for Opus 4.8 on the latter benchmark.

Questions & Answers

Q: What can Claude Opus 5 build from a prompt?

Claude Opus 5 can build interactive browser applications that combine generated visuals, controls, simulation logic, sound, and game-like behavior. The creator demonstrated a 3D art gallery, orbital launch game, cloth simulation, fractal generator, cellular automaton, ecosystem model, quadcopter simulator, pixel editor, sneaker configurator, historical population graph, and wrecking-ball simulation, all produced with very little manual effort.

Q: How good is Claude Opus 5 at creating simulations?

Claude Opus 5 appeared particularly capable at producing accessible, interactive simulations. Its examples covered orbital trajectories, wind-driven cloth, growing cells, fluid-like materials, predator-prey dynamics, drone flight, pendulum motion, and destructible walls. The demonstrations looked responsive and engaging, although the creator cautioned that convincing behavior does not prove that the underlying physics, such as fabric dynamics, is completely accurate.

Q: Is Claude Opus 5 cheaper than Fable 5?

Claude Opus 5 was cheaper in the specific website-generation comparison presented. The Airlines website reportedly cost 69 cents to create with Opus 5, while the Fable 5 version cost 94 cents. The creator also considered the Opus result visually stronger and more sophisticated, but these figures describe the showcased task rather than every possible workload or usage pattern.

Q: How does Claude Opus 5 compare with Fable 5?

The creator described Claude Opus 5 as slightly more intelligent than Fable 5 at perhaps half the cost for the kinds of tasks tested. In a direct website comparison, the Opus version used custom SVG resources and a more elaborate sliding presentation, while the Fable version appeared simpler. The reported agentic terminal coding benchmark also placed Opus 5 ahead of Fable 5.

Q: What benchmark scores did Claude Opus 5 achieve?

The transcript reports that Claude Opus 5 scored 1861 on GDPval-AA V2 for knowledge work and 30.2% on ARC-AGI 3 for novel problem solving. Opus 4.8 reportedly scored 1.5% on ARC-AGI 3. The creator also states that Opus 5 led the referenced models on agentic terminal coding, including Fable 5, Opus 4.8, and GPT 5.6 Soul.

Q: Can Claude Opus 5 create a functional game?

Claude Opus 5 can create experiences that function like small games, based on the examples shown. The 3D gallery supported navigation and collectible-style cataloging, the orbital project combined instruction with a launch challenge, and the quadcopter simulator included flight controls, sound, a procedurally generated environment, and helipad landing. These examples went beyond static pages by responding continuously to user input.

Q: What mistakes did Claude Opus 5 make during testing?

Claude Opus 5 produced several visible imperfections during the demonstrations. The creator noticed an apparent design mistake in the upper-right area of the double-pendulum application. Reducing the pixel editor's canvas size also produced an unhappy-looking result, and the generated sneaker was not always clearly recognizable as a shoe. These issues show that its applications still require inspection and refinement.

Q: Can Claude Opus 5 create useful design tools?

Claude Opus 5 created a pixel editor with drawing, erasing, undo, onion-skin functionality, layers, multiple animation frames, and canvas resizing. It also produced a sneaker configurator that allowed users to select sections, change colors and materials, rotate the object, and save a render. The visual execution was imperfect, but both examples demonstrated substantial application logic and interactive controls.

Summary & Key Takeaways

• Claude Opus 5 generated a broad collection of interactive browser experiences, including a 3D art gallery, orbital launch game, cloth simulation, fractal generator, cellular automaton, predator-prey ecosystem, quadcopter simulator, pixel editor, product configurator, population visualization, and wrecking-ball simulator. The creator emphasized their visual quality, interactivity, and rapid production.

• The demonstrations suggest that Opus 5 can combine programming, interface design, generated graphics, sound, physics-like behavior, and simulation logic within complete applications. Some outputs were imperfect, including a double-pendulum design error and an unclear sneaker model, but most examples remained functional enough for the creator to explore, manipulate, and assess directly.

• The creator characterized Opus 5 as slightly more intelligent than Fable 5 at roughly half its cost. One showcased website reportedly cost 69 cents with Opus 5 versus 94 cents with Fable 5, and the creator preferred the Opus result. Reported benchmarks also showed substantial gains over Opus 4.8.