GPT-6.1 Terra (Max) Scores 48 Enhances Artificial Analysis
agents benchmarks reasoning
| Source: HN | Original article
GPT-6.1 Terra (Max) earned a 48 on the Artificial Analysis Intelligence Index, a composite benchmark measuring reasoning and knowledge performance.
OpenAI’s flagship model, GPT‑6.1 Terra (Max), has been given a 48‑point rating on Artificial Analysis’s Intelligence Index, the latest composite benchmark that measures reasoning, knowledge, mathematics and coding. The score places Terra (Max) 25th out of 221 evaluated models, above the class median of 28 but well behind the top‑scoring GPT‑6.1 Sol, which posted 54 on the same Coding Agent Index v2.0 and 49.1 % on FrontierCode 1.1 at maximum effort.
The rating is the first public performance snapshot since OpenAI launched the Terra line on Oct. 27, when the company highlighted the model’s “fast and capable computer use at scale” and its $1 per million input‑token, $5 per million output‑token pricing. By situating Terra (Max) within a broader competitive field, the Artificial Analysis score gives enterprises a clearer sense of the model’s relative strength and cost‑effectiveness, especially as developers weigh Terra against its sibling Sol and the lower‑scoring Luna variant.
The result also underscores how OpenAI’s tiered approach is shaping the market. While Terra (Max) outperforms the median, the six‑point gap to Sol suggests room for optimisation in high‑end coding and reasoning tasks. Observers will be watching whether OpenAI releases firmware updates or fine‑tuned variants that can close that gap, and how rival providers respond with their own benchmark releases.
Future updates from Artificial Analysis, particularly any revisions to the Intelligence Index methodology or new versioned tests, will be key indicators of whether Terra (Max) can climb the rankings and justify its pricing in an increasingly performance‑driven AI landscape.
Sources
Back to AIPULSEN