OpenAI Ships GPT-6.1 Terra at Half the Price of Its Predecessor

OpenAI has launched GPT-6.1 Terra, a mid-tier reasoning model it prices at 50% below GPT-5.6 Terra while claiming better accuracy on professional tasks, visual reasoning, and computer use. The model replaces GPT-5.6 Terra and sits between the cheaper GPT-6 Luna and the higher-capability GPT-6.1 Sol and GPT-6 Astra in the current lineup.

The pitch is straightforward: near-frontier capability without the frontier invoice. "We're introducing GPT-6.1 Terra, a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing," OpenAI said in its announcement. Target workloads include drafting proposals, interpreting technical drawings, filling out forms, and editing videos.

Pareekh Jain, CEO at EIIRTrend & Pareekh Consulting, framed the positioning bluntly. "Enterprises need AI models that balance intelligence, speed and cost. Frontier models are expensive for routine tasks, while smaller models may lack sufficient capabilities. Terra addresses this gap, competing with Anthropic's Claude Sonnet and Google's Gemini Flash models."

What do the benchmarks actually show?

On Agents' Last Exam, a benchmark of professional tasks, GPT-6.1 Terra scored 60.0% at max effort, at a cost of $1.01 per task. The comparison stack:

• GPT-5.6 Terra: 50.7% at $3.58 per task
• GPT-6.1 Sol: 58.6% at $2.00 per task
• GPT-6 Astra: 59.3% at $7.24 per task
• Claude Opus 5.5: 63.2% at $8.82 per task

Terra beats Sol on score and cost simultaneously. That alone will reshape a few procurement shortlists.

Visual reasoning on MMMU-Pro tells a similar story. GPT-6.1 Terra hit 85.5% at max effort for $0.012 per task, against 81.7% at $0.079 for GPT-5.6 Terra. GPT-6 Luna scored 79.9% ($0.0021), GPT-6.1 Sol reached 86.3% ($0.013), and GPT-6 Astra led at 86.7% ($0.065). OpenAI offered no competitor comparisons on this benchmark.

The computer-use story has a catch. On OSWorld 2.1, which measures how agents navigate desktop applications through multi-step tasks, Terra scored 74.2% at $1.19 per task — a large step up from GPT-5.6 Terra's 57.1% at $4.76. Anthropic's Claude Sonnet 5.5 still leads the category at 83.9%, though at $5.74 per task.

On reliability, OpenAI says its internal benchmark of error-inducing prompts shows the error rate at low reasoning effort falling from 53.6% on GPT-5.6 Terra to 16.7% on the new model. At maximum effort, errors dropped 18 percentage points to 4.3%, nearly matching GPT-6 Astra and GPT-6.1 Sol. On cybersecurity, Terra meets the Critical threshold under OpenAI's Preparedness Framework and keeps the safeguards built for GPT-6 Astra.

What does this change for workload placement?

Jain draws a practical line between the two siblings. "Terra is ideal for high-volume tasks such as invoice processing, document analysis, application automation and routine coding. It improves on GPT-5.6 Terra in speed and accuracy, while GPT-6.1 Sol is better suited for complex reasoning, research, and software engineering."

He also cautions that benchmarks alone are insufficient. Enterprises should run Terra against their own workloads, measuring task completion rates, accuracy, execution time, cost per workflow, human intervention, and compliance with security requirements. A vendor scoreboard is not a deployment plan.

What does it cost, and where is it available?

GPT-6.1 Terra is available now to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, and through the OpenAI API as gpt-6.1-terra. Token pricing per million:

• GPT-6.1 Terra: $1 input, $5 output, $0.05 cached input
• GPT-6 Luna: $0.10 input, $0.50 output, $0.01 cached input
• GPT-6.1 Sol: $2 input, $10 output, $0.10 cached input
• GPT-6 Astra: $10 input, $50 output, $1 cached input
• Claude Sonnet 5.5: $2 input, $10 output, $0.10 cached

Terra undercuts Claude Sonnet 5.5 by half on both input and output rates.

Cheaper tokens, unchanged arithmetic?

Sanchit Vir Gogia, chief analyst at Greyhound Research, warns against reading the price cut as a total-cost story. "Cheaper inference improves one component of agent economics. Terra's base rates are half Sol's, but review and integration still require funding," he said.

His math deserves a place in any budget spreadsheet. "Token processing is fuel with a higher operating cost. A CFO relieved by cheaper fuel needs to establish what happens to acceptance and review demand. In the illustrative coding workload, 5.4 additional review seconds erase the inference saving. At scale, the relevant purchase is completed work that remains correct and authorized."

The verdict: Terra halves the input bill and lifts most scores, but Anthropic still owns the computer-use crown, and review overhead can quietly claw back the savings. Pilot it, measure the whole workflow, then commit.