Which model is worth using
1 Oct 2026
The same snapshot as the board, asked as a question. Change the effort and the recommendation moves with it.
Strongest you can call
54 on the index · $1.82 per measured task
Long coding sessions and knowledge work, when you want the highest generally available score on this index. High effort is the setting to buy. Max effort adds a few points and about triples the measured task cost.
Best price, closed model
50 on the index · $0.32 per measured task
The default OpenAI model. Coding and agent work at a price you can leave on. High effort is 50 on the index at $0.32 a measured task. Max is 52 at $0.72.
Best open weights
46 on the index · $0.13 per measured task Closest published setting, not High.
The open model to try before you pay frontier prices. Index 46, the same band as Grok 4.7, at $0.13 per measured task and a 1M window.
Daily coding
Index 48 or better, measured task cost at or under $2, then the most index points per dollar.
Which generally available model stays near the top of the index without a max-effort bill?
The default OpenAI model. Coding and agent work at a price you can leave on. High effort is 50 on the index at $0.32 a measured task. Max is 52 at $0.72.
Someone is waiting
Highest output speed among models at index 35 or better. Speed is tokens per second after the model is writing, not time-to-first-token.
Which capable model starts producing tokens the fastest?
Multimodal product work you need to ship now: image, audio, and video in, with a 1M window and very fast output. High effort is 41 on the index at about 236 tokens per second.
A long document
Largest window among models at index 40 or better. Fill it only after you read the long-context surcharge. Some labs reprice the whole request.
Which strong model can take the most context?
The hardest OpenAI jobs: long tool use, computer use, and work where you want the flagship specifically. High effort is close to max on the index and about half the measured task cost.