2024 Review: Large AI Models and Declining Service Costs

2024 Annual Review of Large Models: Disruption and Declining Costs

In the rapidly evolving field of artificial intelligence in 2024, large model technology has witnessed revolutionary breakthroughs. The emergence of OpenAI's Sora, alongside several tech giants launching substantial products, has disrupted the market monopoly held by GPT-4. This year's advancements reflect dual breakthroughs in both performance and pricing.

Increased Competition and Model Advancements

The AI landscape has become increasingly competitive this year, with 18 companies releasing over 70 new models. Many of these models are reported to surpass the performance of GPT-4. Notably, Google's Gemini 1.5 Pro has introduced a new era of multimodal interaction, showcasing groundbreaking capabilities in long text processing and video understanding. Additionally, products like the Claude 3 series and Meta's Llama 3.3 have gained traction among users, particularly the Claude 3.5 Sonnet, which has demonstrated exceptional performance in various evaluations.

Image Source Note: Image generated by AI, image authorized by service provider Midjourney

Reduction of Technical Barriers

A significant breakthrough this year is the reduction of technical barriers. Models previously requiring substantial computational power, such as those at the GPT-4 level, can now operate on standard laptops. For example, the Qwen 2.5-Coder-32B runs efficiently on an M2 MacBook Pro, providing unprecedented convenience for developers and everyday users.

Declining Prices Amidst Competition

The intensified competition in the market has led to a significant decrease in service costs. Google's Gemini 1.5 Flash, priced at only $0.075 per million tokens, exemplifies this trend. This reduction not only reflects the fierce market competition but also highlights substantial improvements in model efficiency. Additionally, the energy consumption per call has been significantly reduced, easing the environmental burden while maintaining efficient service delivery.

Implications for the Future of AI Technology

The advancements in large models throughout 2024 demonstrate a remarkable dual breakthrough in performance enhancement and cost reduction. These improvements are expanding application boundaries, allowing ordinary users to access top-tier AI technology with a lower entry threshold. This series of advancements indicates that AI technology is entering a more open and inclusive new era.

Key Points

1. The AI landscape has seen increased competition with over 70 new models launched.
2. Google's Gemini 1.5 Pro introduces advanced multimodal capabilities.
3. Significant reductions in service costs are making AI technology more accessible.