Key insights
- Jefferies has lowered its view on the AI sector due to intensifying competition and a widening US-China intelligence gap. While Meta's AI model Muse Spark 1.3 improved its ranking, the overall trend shows increased pricing pressure and strategic shifts, with OpenAI targeting lower-end markets. Western enterprises continue to favor US models for complex tasks due to lower hallucination rates.
Investing.com - Jefferies highlights intensifying competition in the AI sector as several companies leapfrog in intelligence rankings and pricing pressure mounts, with the US-China intelligence gap widening to 20% from 12% between August and September.
In September, US models occupied the top four spots in Artificial Analysis' intelligence ranking, with Anthropic retaining the lead with Claude Opus 5.5, followed by OpenAI's GPT-6 Astra. Meta (NASDAQ: META) jumped to third place with Muse Spark 1.3, while Chinese models occupied positions five through nine, led by Xiaomi's MiMo V2.6 Pro, Qwen3.8 Max, and GLM-5.3. Previous laggards including Meta, Xiaomi, and StepFun largely jumped in the rankings, and four new models entered the ranking in September.
The average API price gap between US and Chinese frontier models widened to 70% in September from 60% in August, partly because OpenAI launched GPT-6 Astra at a 150% premium to GPT-5.6 Sol. Anthropic priced Opus 5.5 24% below Opus 5.0, while OpenAI cut GPT-6 Sol and Luna prices by 50% and 56% versus their GPT-5.6 equivalents. GPT-6 Luna is now 74% cheaper than DS V4.1 Flash, suggesting OpenAI is pushing into mid- to low-end segments dominated by Chinese models.
Jefferies notes that Chinese AI players can partly offset compute constraints through more intensive reinforcement learning, boosting intelligence without drastically increasing model size, but this approach has drawbacks. Aggressive reinforcement learning can weaken performance in areas without standardized answers such as creative writing and business analysis, increase hallucinations due to overconfidence, raise token consumption and costs, and lengthen inference time. The firm recently hosted AI expert William Fong, who observed that Western enterprises rely more heavily on frontier US models for complex task planning and result verification given their lower hallucination rates.
World models, which aim to understand and simulate characteristics of the physical world with potential applications in autonomous driving, robotics, gaming, and entertainment, could be the next development in AI but remain at a very early stage. These models would require much more compute and memory, making return on investment an even bigger challenge. AMD announced Wednesday its $8.2 billion acquisition of prominent world-model startup World Labs.