How to Choose Between Claude 5 and GPT-5.6 Model Tiers
Claude and ChatGPT recently released major platform updates, introducing model tiers like Opus 5, Sonnet 5, and GPT-5.6 Sol, Terra, and Luna. These updates allow teams to match the model to the specific task, improving efficiency and capabilities.
Opus 5, Sonnet 5, and GPT-5.6
Claude and ChatGPT both shipped major platform updates over the last six weeks. Anthropic launched Opus 5 on July 24, delivering a one-million-token context window, up to 128K output, adaptive thinking on by default, and a five-level effort setting. Sonnet 5 arrived earlier on June 30, offering near-Opus coding quality at Sonnet prices.
OpenAI released GPT-5.6 broadly on July 9 in three flavors: Sol, Terra, and Luna. Sol is the most powerful, Luna is built for speed, and Terra balances the two. On July 30, OpenAI cut the price of Luna by 80% and Terra by 20%.
Pick a Model Tier for the Specific Task
The most useful change is not raw capability. It is that both platforms now give you a tier for the task. You can choose fast and cheap models for high-volume routine work, and heavy, deliberate models for the analysis you would otherwise assign to a person for a week.
If your team runs everything through one default model, you are either overpaying for simple work or underpowering the hard work.
What a Million-Token Context Window Means
A million tokens is an abstract number until you translate it into daily work. It means you can hand the model an entire year of board decks, a full contract portfolio, or every customer survey from last quarter, and ask questions across all of it at once.
There is no chunking, no summarizing first, and no losing the thread halfway through.
Fable 5 Returns and Atlas Browser Notice
Following up on the June update regarding the US government ordering Anthropic's Fable 5 offline two days after launch, the disruption is over. The Commerce Department lifted the export controls on June 30, and Anthropic restored the model globally on July 1, after eighteen days offline. The precedent is still worth watching.
As a practical heads-up, if anyone on your team uses the Atlas browser, it stops working on August 9. Move your bookmarks and data before then, as they will not transfer automatically.
Bring Practical AI to Your Team
Book a custom AI workshop or keynote to help your team choose the right models and build practical workflows. Let us close the gap between having the tools and using them.
Book a 30-Minute Call
Jon Lakefish
Jon Lakefish is one of North America's most sought-after AI keynote speakers and workshop facilitators and is available for events worldwide. He helps business teams cut through the hype and leave with practical AI tools they can use immediately.