| 1. Claude Opus 5 (High) |
| 2. Claude Opus 5 (Max) |
| 3. Claude Fable 5 (High) |
| 4. GPT 5.6 Sol (xHigh) |
| 5. Claude Opus 4.8 (High) |
The BriefOpenAI dropped the price of its best model by more than a fifth and promised to hold it there through November 21. If you build on GPT-5.6 Sol, the same work now costs about four fifths of what it did last week, with no code change on your side. The reason is competition: Anthropic, Google and the Chinese open-weight labs are all shipping fast, and price is the lever that moves fastest. If you priced an AI feature months ago and shelved it as too expensive, this is the week to run those numbers again.
Level UpHugging Face just published the numbers on who is actually releasing open models, and one finding should change how you pick one. Tiny models do the real work: everything under 1 billion parameters accounts for 83 percent of all downloads ever, while models above 100 billion take just 1 percent. This week, stop reaching for the biggest model you can afford and test the smallest one that clears your bar. Read the report first, then pick one job you already automate and try it on a small model. State of Open Models, Summer 2026: who ships, who gets downloaded, and the size nobody expected