OpenAI releases GPT-6.1 Terra with advanced computer control capabilities

On October 26, 2026, OpenAI launched GPT-6.1 Terra, a mid-tier AI model designed to deliver high performance at reduced operational costs. Priced at $1 per million input tokens and $5 per million output tokens—half the API cost of GPT-5.6 Terra—the model achieves a score of 48 on the Artificial Analysis Intelligence Index. GPT-6.1 Terra offers three times faster execution in computer control tasks, solves 50.3% of Terminal-Bench 4.0 coding problems, and maintains advanced cybersecurity defenses.

Model release and availability

• ▪OpenAI released GPT-6.1 Terra as a mid-tier AI model balancing speed, capability, and cost
• ▪GPT-6.1 Terra is available to Plus, Pro, Enterprise, and Edu subscribers in ChatGPT Work and Codex, as well as via API

API pricing and lineup

• ▪OpenAI set GPT-6.1 Terra API pricing at $1 per million input tokens and $5 per million output tokens, half the price of GPT-5.6 Terra and GPT-6.1 Sol
• ▪OpenAI's full model lineup spans four cost tiers: GPT-6 Luna at $0.10/$0.50 per million input/output tokens, GPT-6.1 Terra at $1/$5, GPT-6.1 Sol at $2/$10, and GPT-6 Astra at $10/$50
• ▪OpenAI reduced cached input token prices for GPT-6.1 Terra to $0.05 per million tokens, a quarter of GPT-5.6 Terra's previous rate

Computer control and agent performance

• ▪OpenAI reported that GPT-6.1 Terra operates more than three times faster than both GPT-6 Astra and GPT-6.1 Sol while executing computer control tasks
• ▪On OpenAI's desktop agent benchmark OSWorld 2.1, GPT-6.1 Terra outperformed GPT-5.6 Terra's best medium-effort result at lower cost
• ▪On the Agents' Last Exam benchmark, OpenAI's GPT-6.1 Terra matched GPT-6 Astra's computer control performance at one-tenth of GPT-6 Astra's token price

Coding benchmark performance

• ▪On Terminal-Bench-Science 0.1, GPT-6.1 Terra outperformed GPT-6.1 Sol at lower cost and scored within five points of Claude Opus 5.5 at under a fifth of the estimated cost
• ▪On the Coding Agent Index v2.0 benchmark, GPT-6.1 Terra scored 48 points, placing it 17 points above GPT-6 Luna and six points below GPT-6.1 Sol
• ▪On Artificial Analysis's Terminal-Bench 4.0 coding benchmark, GPT-6.1 Terra solved 50.3 percent of tasks at maximum reasoning effort, up from 21.5 percent for GPT-5.6 Terra

Artificial Analysis benchmark results

• ▪On the Artificial Analysis Intelligence Index, GPT-6.1 Terra scored 48 points, placing it between GPT-5.6 Terra at 42 points and GPT-6.1 Sol at 52 points
• ▪On Artificial Analysis's private AA-Omniscience benchmark, GPT-6.1 Terra resorted to guessing 38 percent of the time when lacking an answer, down from 88 percent for GPT-5.6 Terra
• ▪Artificial Analysis reported that at maximum reasoning effort, GPT-6.1 Terra gains 6 points on the Artificial Analysis Intelligence Index over GPT-5.6 Terra while costing $0.40 per task compared to $1.40

Safety and security protections

• ▪GPT-6.1 Terra met OpenAI's Critical cybersecurity threshold under OpenAI's Preparedness Framework and inherited safety safeguards from GPT-6 Astra
• ▪Against direct prompt injection attacks, OpenAI's GPT-6.1 Terra achieved a 99.99 percent defense rate, matching GPT-6 Astra and GPT-6.1 Sol
• ▪OpenAI reported that GPT-6.1 Terra increased its defense rate against indirect prompt injection attacks to 98.8 percent, compared to 96.2 percent for GPT-5.6 Terra