Get your free personalized podcast brief

We scan new podcasts and send you the top 5 insights daily.

George Kurtz explained that running general frontier models for security remediation costs roughly ten dollars per instance, whereas CrowdStrike's post-trained model costs three cents. Feeding a decade of proprietary incident response data into smaller models with customized software harnesses yields superior security performance without overpaying for general-purpose reasoning.

Related Insights

For specialized, high-stakes tasks like insurance underwriting, enterprises will favor smaller, on-prem models fine-tuned on proprietary data. These models can be faster, more accurate, and more secure than general-purpose frontier models, creating a lasting market for custom AI solutions.

For most enterprise tasks, massive frontier models are overkill—a "bazooka to kill a fly." Smaller, domain-specific models are often more accurate for targeted use cases, significantly cheaper to run, and more secure. They focus on being the "best-in-class employee" for a specific task, not a generalist.

Nikesh Arora argues that using large, token-based AI models to inspect massive data flows in cybersecurity is cost-prohibitive. The viable strategy is deploying specialized small language models (SLMs) for widespread monitoring, where marginal cost is zero, reserving expensive models only for confirmed threats.

The economics of AI security requires a tiered approach. The optimal strategy involves using low-cost, domain-adapted open models ("drones") for constant monitoring across the entire environment, while reserving expensive frontier models ("battleships") for selectively hunting novel threats, balancing cost and coverage.

Decagon found that fine-tuning smaller, specialized open-source models led to better performance, lower latency, and reduced costs compared to using general-purpose frontier models for specific enterprise workflows. This contradicts the common "smart but expensive vs. dumb but cheap" tradeoff.

Enterprise workloads rarely require broad superintelligence; instead, fine-tuned nine-billion-parameter open-weight models achieve superior accuracy on specific tasks like fraud detection. These smaller models run faster, cost far less compute, and allow corporations to preserve AI sovereignty without paying exorbitant frontier model fees.

Analogous to just-in-time compilers that emit machine code on the fly, systems could use general foundation models to autonomously train narrow, domain-specific replacement models. Amjad Masad points out that using frontier models for simple tasks is like nuking a butterfly; dynamically training small replacement models cuts operational cost and dramatically limits vulnerability to prompt injection because their capabilities are strictly bounded.

Instead of relying on expensive, omni-purpose frontier models, companies can achieve better performance and lower costs. By creating a Reinforcement Learning (RL) environment specific to their application (e.g., a code editor), they can train smaller, specialized open-source models to excel at a fraction of the cost.

Microsoft's strategy lets companies customize proprietary models for specific tasks, achieving near-frontier performance at a fraction of the cost. This 'controlled tuning' approach is a powerful alternative to using expensive general models or relying on potentially inaccessible open-source options from abroad.

Fine-tuning, once dismissed as obsolete due to powerful foundation models, is making a comeback. Faced with expensive pay-as-you-go APIs for long-running agents and geopolitical supply risks, companies are returning to fine-tuning smaller, self-hosted models to gain cost control and operational resilience.