GPT-6.1 Terra: release report
OpenAI released GPT-6.1 Terra on October 26, 2026. Specs, price, how it plays chess and how it solves sudoku, with more added as evidence arrives.
The release
OpenAI released GPT-6.1 Terra on October 26, 2026. Read the announcement.
- Takes unknown. Returns unknown.
- Not a reasoning model.
- Closed weights.
OpenAI's release before this was GPT-6.1 Sol on September 29, 2026, 27 days earlier. Over the past year it has shipped a model every 12 days on median (24 releases).
Price
$1.00 per million input tokens and $5.00 per million output tokens on OpenAI's API. Cached input is $0.050. GPT-5.6 Terra, the model it follows, costs $2.00 and $12.00.
Our chess benchmark
Rated 1,100 (95% interval 945 to 1,254) over 25 games against calibrated Stockfish, number 8 of 15 models we've tested. Standard chess 1,164, Chess960 1,039.
In standard chess: 86.0 accuracy in open positions, 6.0 blunders per 100 moves, 3.0% of answers were illegal moves.
Each game took 3.9 minutes of model time. At the lab's list price its tokens would cost $0.11 a game.
Our sudoku benchmark
Scored 7 out of 100 (95% interval 0 to 17), number 8 of 9 models we've tested. It solved 3 of 42 grids and filled 10.2% of blank cells correctly.
By tier, from singles to grids no human technique finishes: Cozy 3 of 6, Gentle 0 of 6, Steady 0 of 6, Tricky 0 of 6, Devious 0 of 6, Expert 0 of 6, Beyond 0 of 6.
It said it was 12% sure on average; 2 wrong answers came with 90% confidence or more; 2 finished grids repeated a digit.
Each grid took 26 seconds and 1,849 output tokens on average. At the lab's list price its tokens would cost $0.015 a grid.
Each section was written on the date shown and is not revised afterwards. Specifications from Models.dev, prices from provider listings, capability scores from Epoch AI under CC BY 4.0, chess and sudoku results from Glimmer Index Bench.