Text and chat AI models: the 228 that write and converse

Text generation is the baseline every language model shares, which is why 228 of the 239 models in this rating qualify - and why price, not capability, is usually what decides. Output rates span two orders of magnitude: $0.25 per million tokens at the budget end, $50 at the frontier with Claude Fable 5.1 and GPT-6 Astra.

For ordinary writing, summarising and conversation, the mid tier is where the value sits. Claude Sonnet 5 at $10 and GPT-5.6 Terra at $12 handle most professional work, and Gemini 3.8 Flash does it for $3.75 while promotional pricing lasts. Reserve the frontier models for work where a wrong answer is expensive.

• Decide whether you need reasoning at all — plain chat models answer faster and cheaper.
• Match the context window to your longest realistic input, not to the largest number on offer.
• Check whether the vendor offers batch and caching; both change the economics more than the headline rate.
• If you serve non-English users, test tokenisation cost on real text before committing.