I made Gemini 4 Argon, Fable 5.1, and GPT-6 Astra build the same audio tool, and the best-looking one wasn't the best-sounding one
A recent comparison of AI models Gemini 4 Argon, Fable 5.1, and GPT-6 Astra in building a drum machine found that Google's Gemini 4 Argon produced the most visually distinct design. Despite Bloomberg reporting some employee concerns about its front-end capabilities, Argon's creation demonstrated significant personality and cultural understanding, though its audio output was described as thin.
Google's Gemini 4 Argon, which the company announced with strong benchmark scores but which some Google employees with access reportedly felt struggled with front-end design, delivered the most aesthetically compelling drum machine in a recent test. The model named its creation the 'TR-86 Rhythm Composer,' echoing classic hardware, and incorporated visual elements like wood-paneled sides and a red LED tempo display. Its presets, named after genres such as '909 house' and 'boom bap,' suggested a notable understanding of the cultural context surrounding such instruments.
While Argon's design leaned towards user-friendly sliders over authentic knobs and had minor layout bugs when resized, its visual flair was prominent. However, its audio performance was less impressive. The synthesized sounds were stable but described as 'thin,' with the kick drum lacking low end and the hats sounding 'fizzy.' This indicates a potential gap between a model's ability to generate visually rich, culturally informed code and its capacity to produce functionally high-quality audio, especially when it cannot directly 'hear' the Web Audio API output it creates.
In contrast, Anthropic's Fable 5.1, described as shining in 'steps and tool use,' approached the task methodically. It created a detailed task list, wrote files directly, and self-checked for console errors, taking five minutes just for planning. Fable also exceeded the brief by adding extra tracks, demonstrating a thorough, process-driven approach that prioritizes functional completeness, distinct from Argon's design-centric output.
Share this article
Related reading
6 stories
Google rolling out Gemini app for CarPlay
Google is rolling out its Gemini app for CarPlay, showing how these models are being integrated into everyday tools.

Vietnamese Travellers Can Now Scan NETS QR Codes to Pay in Singapore

Weixin Pay Sees 84% Rise in Overseas E-Wallet Transactions

BPC Pairs SmartVista with Huawei for Payment Modernisation

