Episode 71 · Season 1
Claude vs GPT-5.6: How to Choose the Model Behind Your AI Agents
Every agent you build runs on a large language model, and that choice quietly decides whether your AI program pays for itself or drains the budget. Cirrius Solutions leaders Tim Harting and Gavin Franklin compare the Anthropic and OpenAI lineups tier by tier, from model families to subscription plans to what a task really costs to run.
What you’ll learn:
- How the two model families mirror each other across four tiers, from Haiku and Luna for cheap automation up to Fable 5 and Sol Ultra at the frontier
- What each subscription actually buys, from free and $8 entry plans through $100 and $200 credit pools, team seats, and usage-based enterprise deals
- Why cost per task, not printed per-token price, is the number to watch, using live Artificial Analysis benchmarks
Who it’s for:
Business leaders, IT directors, and CIOs choosing the foundation models and plans to build their agentic platform on, and anyone trying to budget AI spend for a team.
Guest + credibility:
Tim Harting is the Chief Operating Officer of Cirrius Solutions and a recurring co-host of the series. Gavin Franklin is Cirrius’s Head of Custom Development, where he builds agentic systems for enterprise clients on these same model families.
Primary CTA:
Follow Cirrius Talk: Agentic AI so the next pricing shake-up never catches your budget off guard.
Recommended Links & Resources:
We cover
- Navigating AI Models and Costs
- Understanding Large Language Models
- Evaluating Model Effort and Complexity
- Comparing Model Families and Costs
- Subscription Tiers and Their Implications
- Real-World Cost Analysis of AI Models
Chapters
- 0:44Navigating AI Models and Costs
- 3:25Understanding Large Language Models
- 6:05Evaluating Model Effort and Complexity
- 8:33Comparing Model Families and Costs
- 11:17Subscription Tiers and Their Implications
- 13:35Real-World Cost Analysis of AI Models
- 15:59Strategizing AI Usage in Teams
- 18:49Final Thoughts on AI Model Selection
Resources and downloads
The artifacts Greg’s team built for this episode, free to download.
Anthropic vs OpenAI: Which LLM Should Power Your AI Agents?
A side-by-side comparison of Anthropic’s Claude 5 family and OpenAI’s GPT-5.6 family across four tiers of reasoning depth, from fast and cheap classification up to frontier agentic work. Includes API output pricing per million tokens at each tier, a decision matrix mapping common tasks such as support triage, production workloads, and long-horizon coding to a recommended model from each vendor, a walkthrough of what the subscription tiers actually include at matching price points, and the roughly 150-seat inflection where flat-rate pricing ends for both vendors and they diverge in opposite directions.
Related episodes
Agentic AIIs Your Team’s Claude Code Setup the Wild West? Plugins Fix That
Agentic AISkills vs Commands: When to Use Each in Your Agentic OS
Agentic AISignal, Sort, Store: The Discipline Behind Smart Agents
The Military MinuteCamp Twin Lakes: The $50 Weekend Veteran Families Call a LifelineKate Lipton, Recreational Therapist and Military Family Programs Lead
Salesforce at Scale48% of Salesforce Storage Wasted by One Ex-Admin – AI Found It in 2 HoursNick Stump, Senior ConsultantTalk to us about AI agents
Keep going: See how we build AI agents for real businesses, Talk to us about a custom AI agent, More from the Cirrius blog