GPT-5.6 Sol, Terra and Luna: which model is good at what?
Sol is for deep difficult work, Terra for balanced daily production, and Luna for speed and volume. Where can you use each one, and when should you choose it?
Which AI model fits which task?
I compare AI models and coding agents through real use cases, then explain which tool, model and setting makes sense for each task.
Featured guide
Read the guide →A practical guide to model choice.
Compare Sol, Terra and Luna by task type, then choose the model that fits your work.
Articles
All articles →Guides that make choices clearer.
Practical guides to model selection, coding agents, reasoning settings and output verification.
MiniMax M3.1 Flash launches: Cheap coding, but no model weights yet
The October 19 launch offers a one-million-token context and $0.15/$0.60 API rates. Vendor speed claims differ from independent measurements; here's who should pilot it.
Anthropic let METR inside: Independent AI evaluation depends on access and publication rights
Anthropic gave six METR staff internal access and publication rights. Why is the arrangement substantial, and why is it still not an audit?
Gemini 4 Argon is now available: Should it move from your watchlist to your pilot queue?
Google’s new model is no longer limited to selected security partners. The useful question is not whether Argon is the best, but which workloads justify a measured pilot.