EXPLORINEER
en

Archive — US — techcrunch

Apple’s upgraded AI models underwhelm on performance

Apple's latest AI model updates for iOS and macOS have been found to underperform compared to competitors like OpenAI and Google, with human testers rating them lower, despite enhancements in efficiency and tool-use. This highlights ongoing struggles in Apple's AI development amid customer dissatisfaction and a delayed Siri upgrade.

Key points

  • Apple updated AI models for AppleIntelligence features across iOS and macOS.
  • Benchmarks show that Apple's models underperform compared to older models from OpenAI and others.
  • Human testers rated the new 'Apple On-Device' model similarly to competitors but not better.
  • The 'Apple Server' model was rated lower than OpenAI's GPT-4o.
  • In image analysis tests, Apple's models were outperformed by Meta's Llama 4 Scout.
  • Apple's AI research division has struggled against competitors, leading to delays in promised updates like Siri.
  • Legal issues arose as customers sued Apple over unfulfilled marketing claims on AI features.
  • The new models integrate improved efficiency and can handle around 15 languages.

Why it matters

  • The struggles of Apple's AI capabilities reflect a broader trend of tech companies racing to enhance their AI models.
  • By 2025, competitiveness in AI will likely center on model efficiency and versatility, as seen with advancements from rivals.
  • Apple's challenges illustrate ongoing issues in the tech industry concerning innovation pace and customer expectations.

Despite efforts to upgrade AI capabilities, Apple's latest models have not met performance expectations, emphasizing the challenge of keeping pace with AI advances in a competitive landscape.