Perplexity CEO states that a model with Fable 5 quality will emerge in the next 6 months at 3-4 times lower cost
Perplexity CEO Aravind Srinivas stated that a model with Fable 5 quality will emerge in the next 6 months at 3-4 times lower cost, and a model at Opus 4.8 level could run locally within 12 months, with a probability exceeding 50%.
He emphasized the need to consider these advancements when making future predictions.
Market Mechanism: The enhancement of model capabilities and reduction in costs accelerate the commercialization of AI, driving funds from high-priced frontier models to efficient local/low-cost solutions; under this scenario, applications like Perplexity benefit from better cost-performance infrastructure, putting pressure on traditional cloud reliance models.
Source: Public information
ABAB AI Insight
As the CEO of Perplexity, Aravind Srinivas has been tracking AI search and model advancements for a long time. This prediction continues his optimistic assessment of rapid scaling and efficiency improvements, having previously emphasized the combination of search and reasoning capabilities.
On the capital front, the industry is pushing the cost curve through architecture and hardware optimization, motivated by the desire to expand adoption rates and capture the local/edge computing market. Srinivas's views represent the application layer's expectations for the commercialization of underlying models.
Similar to the efficiency improvements seen in open-source projects like Grok or the Llama series, or historically driven by Moore's Law, this case shows that AI is currently in a phase of accelerated capability-cost dual curve, with localization becoming an important branch.
Structural Judgment: Essentially a technological substitution. Models are migrating from high-cost cloud environments to low-cost local setups, driven by breakthroughs in training efficiency and inference optimization, shifting pricing power from hardware/cloud providers to efficient model developers, expanding AI penetration while compressing high-margin space.
ABAB News · Cognitive Laws
Doubling capabilities, halving costs; to predict the future, first look at the curve.
When top models run locally, cloud monopolies become history.
The faster the technological acceleration, the shorter the business moat; efficiency is the new leverage.