GPT-6.1 Terra: OpenAI’s Half-Price Model for Computer Use and AI Agents

What Launched Today

OpenAI rolled out GPT-6.1 Terra today, upgrading the medium tier of its model lineup (Luna → Terra → Sol → Astra) from GPT-5.6 to GPT-6.1 for the first time. Terra was notably left out of September’s GPT-6 refresh of Sol and Luna — today it catches up, and then some.

OpenAI is positioning Terra as the workhorse model: fast, efficient, and built for “the communication and computer-based tasks that fill the workday, from drafting proposals and interpreting technical drawings to filling out forms and editing videos.” It combines detailed visual reasoning, precise computer use, and clearer, more natural writing — with a 1,050,000-token context window that accepts text and images.

It’s available now in ChatGPT Work and Codex for all Plus, Pro, Enterprise, and Edu users, through the OpenAI API as gpt-6.1-terra, and it’s already listed on OpenRouter — which means your favorite third-party AI tools can adopt it within hours.

The Benchmarks: Frontier Scores at a Fraction of the Cost

The most interesting numbers aren’t raw IQ — they’re how much capability OpenAI preserved while cutting the price:

• Agents’ Last Exam (a benchmark for how well agents use real software at work): Terra at medium reasoning effort scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost, and performs on par with GPT-6 Astra at one-tenth the price per token.
• Error-inducing prompts (OpenAI’s internal reliability benchmark): the error rate at low reasoning effort fell from 53.6% on GPT-5.6 Terra to 16.7% on GPT-6.1 Terra. At maximum effort it drops to 4.3% — nearly matching GPT-6 Astra and GPT-6.1 Sol.
• Against Anthropic: at maximum reasoning effort, OpenAI claims Terra is over three times as token-efficient as Claude Opus 5.5 at one-fourth the cost.

The pattern is consistent: the capability gap between the cheap tier and the flagship tier is collapsing, while the price gap stays wide. That has real consequences for which AI tools you should be paying for.

Why Computer Use Is the Headline Feature

“Computer use” — an AI model that can look at a screen, understand what it sees, and take actions across applications — used to be a premium, experimental capability. Terra makes it a mid-tier commodity. OpenAI specifically calls out use cases like matching product photos to catalog entries, interpreting technical drawings, and filling out forms: repetitive, screen-bound work that fills hours of many jobs.

That’s exactly the workload where agentic AI tools live or die. If you’ve tried browser agents or autonomous workflow tools and found them too slow or too expensive to run at scale, a fast model with GPT-6-class reliability at half the old price changes the math. Long agent sessions burn tokens continuously, so per-token cost and error recovery dominate the real bill — and Terra improves on both axes at once.

On the API side, Terra supports streaming, structured outputs, function calling, file search, and web search, with tool calling handled through the Responses API. In other words, it ships with everything an agent framework needs on day one.

The Price Cut That Changes Tool Economics

The quietest but most consequential detail: GPT-6.1 Terra’s standard API prices are 50% lower than GPT-5.6 Terra — the model it replaces, which was already the budget option. A cheaper, smarter medium tier ripples through the whole AI tool market:

• AI coding agents like Codex, Cursor, and GitHub Copilot can route routine edits and boilerplate work to Terra and reserve flagship models for architecture and tricky debugging — cutting subscription costs or margins under them.
• Automation platforms running high-volume workflows (form filling, document processing, catalog matching) get the biggest win, because their costs scale linearly with volume.
• Competitive pressure: an efficient model that beats Claude Opus 5.5 on token efficiency at a quarter of the cost forces everyone’s pricing down. That’s good for anyone buying AI tools rather than building them.

It also sharpens the strategic picture for everyday users of ChatGPT: the free and Plus experience quietly inherits medium-tier intelligence, while Pro subscribers get Terra’s max-effort reliability for heavier work.

Frequently Asked Questions