GPT-6.1 Terra: intelligence, performance and price analysis
C · Weak signal An Artificial Analysis model page for OpenAI's GPT-6.1 Terra: independent benchmark and price measurements, but a dashboard snapshot with no analysis, method detail or per-benchmark scores in the saved text.
Summary based on the full text·Original: about 5 minutes to read
In short
The Artificial Analysis profile of GPT-6.1 Terra at maximum reasoning effort, an OpenAI proprietary model released on 26 October 2026. It places the model 13th of 59 on the firm's composite Intelligence Index among models in its price class, at a low cost per task. Output speed is above average, but first-token latency is very long and the model is verbose.
Key numbers
- 51score on the Artificial Analysis Intelligence Index v5.0, rank 13 of 59 in its class (class median 38)
- $0.96average cost per Intelligence Index task
- $1.00 / $5.00price per million input and output tokens on OpenAI's API, with a 95% cache discount
- 264.33stime to first token, against a class median of 6.15s
- 99Moutput tokens generated to run the Intelligence Index, against a class median of 71M
As reported by the source.
English·Web page·opens artificialanalysis.ai
What it reports
- Intelligence Index v5.0 score 51; the index combines nine evaluations, including AA-Briefcase, Terminal-Bench 4.0, FrontierCode 2.0, GDP.pdf and HLE-Diamond.
- Price $1.00 per million input tokens and $5.00 per million output tokens (class medians $2.00 and $10.00); $0.73 blended at a 7:2:1 cache-hit/input/output ratio.
- Cost per Intelligence Index task $0.96, rank 12 of 58 on cost.
- Output speed 106.8 tokens per second (class median 84); time to first token 264.33s (median 6.15s).
- 1M-token context window; text and image input, text output; available through 6 API providers.
How they got it
Artificial Analysis runs its own evaluations against the model's first-party API and compares proprietary models with others in the same blended price band. Full methodology sits on separate pages that were not part of the saved copy.
Limits
- The saved page carries headline metrics only; individual benchmark scores did not render in the text.
- Rankings are relative to a price class, not to all models.
- A snapshot on release day; scores and prices on such pages change.
Why it matters for companies
GPT-6.1 Terra is a mid-priced model scoring well above its price class median, which makes it a candidate for high-volume agent and document work where cost per task matters more than top scores. The four-minute first-token latency at max effort rules it out for interactive use at that setting. Firms should test lower reasoning settings and their own tasks before adopting.
Summary based on the full text. Read the original for the full argument and data.Read the original on artificialanalysis.ai
- Tags
- Computer use