DeepSeek's New Model Nearly Matches GPT-6 Astra on Design—at 1.4% of the Cost

What happened

OpenDesign tested 13 AI models on identical design tasks. DeepSeek V4.1 Flash scored one and a half points behind GPT-6 Astra, while costing about 70 times less to run.

Confirmed

Global impact / market context

If a cheaper AI model performs nearly as well, companies may switch to it, cutting their technology spending. This could pressure higher-priced AI providers to lower costs or improve quality to keep customers.

Analyst inference

AI companies often compete on performance and price. A low-cost rival like DeepSeek could reshape how businesses choose AI tools, potentially shifting revenue toward cheaper providers and affecting investor expectations for premium-priced AI products.

Analyst inference

What to watch

1. Watch whether DeepSeek releases more benchmark results showing its model's performance across other tasks, not just design, to see if the cost advantage holds broadly. Confirmed
2. Investors could compare DeepSeek's pricing with GPT-6 Astra's actual subscription or API fees to estimate potential market share shifts in the AI software industry. Proposed
3. Watch if major companies announce adopting DeepSeek for design work, which would signal real demand and possibly reduce spending on higher-priced AI models. Analyst inference

Affected assets

• GPT — Gold Park

Evidence

• decrypt.co