OpenAI / Wikimedia Commons (Public domain)
OpenAI’s GPT-6.1 Terra ranks #14 in Code Arena WebDev with 1,671 points
The mid-tier model beats its predecessor by 151 points while cutting median cost per task by 28%
OpenAI’s newest mid-tier model just walked into one of the most closely watched coding leaderboards and took a seat near the front of the class.
GPT-6.1 Terra, running in its Max configuration, landed at #14 on Arena.ai’s Code Arena WebDev leaderboard with 1,671 points. It also did the job for less money per task than the model it replaces, which is not usually how upgrades work.
The scoreboard and the receipt
OpenAI announced GPT-6.1 Terra on October 26, 2026. By October 27, it held the #14 spot on the WebDev rankings.
That placement is 151 points above GPT-5.6 Terra, its predecessor. The older model now sits at #56.
The cost side is where things get interesting. GPT-6.1 Terra posts a median task cost of $0.40, compared with $0.56 for GPT-5.6 Terra.
That works out to a 28% reduction in median cost per task. More score, smaller bill.
Standard API pricing tells a similar story. OpenAI has set it 50% lower than GPT-5.6 Terra, at $1 per million input tokens and $5 per million output tokens.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Tokens are the chunks of text a model reads and writes, roughly pieces of words. Developers pay by the million, so a halved rate card adds up fast for anyone running heavy workloads.
Speed and terminal work
On Terminal-Bench 4.0 at max effort, GPT-6.1 Terra solved 50.3% of tasks. GPT-5.6 Terra managed 21.5%.
Terminal-Bench measures how well a model handles command-line work, the text-based interface engineers use to run, configure and debug software.
Speed also got attention. On average, GPT-6.1 Terra completes tasks more than three times faster than both GPT-6 Astra and GPT-6.1 Sol.
Where Terra fits in the lineup
OpenAI positions GPT-6.1 Terra as a cost-efficient mid-tier option. Its focus is visual reasoning and routine computer tasks, sitting below GPT-6.1 Sol and GPT-6 Astra in the hierarchy.
The model is available in ChatGPT Work and Codex. Eligible users can also reach it through the API under the name gpt-6.1-terra.
Codex is OpenAI’s coding-focused product, so putting Terra there signals who the company expects to lean on it. Developers doing everyday build-and-fix work are the obvious audience.
The WebDev leaderboard result reinforces that pitch. Code Arena WebDev ranks models on web development tasks, which is exactly the kind of routine-but-valuable work a mid-tier model is meant to absorb.