Updated
Updated · InfoWorld · Jul 13
Enterprises Should Start AI With Cheapest Credible Model, Not GPT-5.6 Bets
Updated
Updated · InfoWorld · Jul 13

Enterprises Should Start AI With Cheapest Credible Model, Not GPT-5.6 Bets

2 articles · Updated · InfoWorld · Jul 13

Summary

  • A senior analyst urged companies to anchor AI strategy to specific jobs, not the latest LLM release, and to begin with the cheapest credible model that can clear a predefined quality bar.
  • OpenAI’s new GPT-5.6 lineup illustrates the point: Terra and Luna trade some capability for lower cost, with Luna nearly matching the prior generation’s peak at less than half the estimated cost.
  • For most enterprise workloads—summarization, classification, document comparison, support assistance—older or smaller models often remain good enough because deployed models do not degrade just because newer ones appear.
  • Frontier upgrades can still matter in coding and other agentic tasks, where a model that completes 80% of a bounded job instead of 50% can change workflows, but revalidation, prompt drift and API costs can erase the gain.
  • The analyst said private evaluation suites built on real company work should decide upgrades, while routing systems should automatically send each task to the lowest-cost model that reliably passes.

Insights

Is your company's reliance on one LLM vendor a failing strategy? Discover why multi-model orchestration is the future of enterprise AI.
Your AI scores high on benchmarks but fails in production. Why are these standard tests a 'total disaster' for enterprise use?
As AI spending is set to hit $2.59 trillion, are you just burning cash? Learn how to slash your company's LLM costs by up to 98%.

The 67% Crash: How "Cheapest Credible" AI Models and Open-Source Are Reshaping Enterprise AI Costs in 2026

Overview

In mid-2026, OpenAI launched the GPT-5.6 family—Sol, Terra, and Luna—after government clearance, marking a new era in premium AI models. Sol stands out by delivering top-tier results in coding, knowledge work, cybersecurity, and scientific research, while also being more efficient and cost-effective than previous models. Its new 'ultra' setting accelerates demanding tasks, and the model family offers options for both high performance and affordability. This release highlights a shift in enterprise AI strategy, where organizations now balance cutting-edge capabilities with operational fit and total cost, rather than just chasing the latest technology.

...