On Oct 26, 2026, OpenAI introduced GPT-6.1 Terra, positioning it as a fast, efficient model that combines detailed visual reasoning and precise computer use with clearer, more natural writing. In its launch post, OpenAI frames Terra as a significant step up from GPT-5.6 Terra for everyday work: drafting proposals, interpreting technical drawings, filling out forms, and editing videos. The headline economic change is pricing—standard API rates are 50% lower than both GPT-5.6 Terra and GPT-6.1 Sol—while in ChatGPT Work and Codex the same usage allowance covers more tasks than with GPT-5.6 Terra and substantially more than with GPT-6.1 Sol or GPT-6 Astra.
What's new
GPT-6.1 Terra targets teams that live in multimodal, tool-using workflows rather than pure code generation. OpenAI emphasizes speed: the model completes tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol in its testing. For product teams, that latency gap matters when agents loop through screenshots, PDFs, and UI automation.
Safety and scope are part of the pitch too. OpenAI says GPT-6.1 Terra meets the Critical threshold in cybersecurity under its Preparedness Framework, inherits protections developed for GPT-6 Astra, is more likely to stay in scope and disclose limitations, and shows strong prompt-injection resistance in its system card. Those claims are about deployment trust, not raw benchmark scores—but they matter for enterprises routing customer data through agents.
Benchmarks (OpenAI's evaluations)
The numbers below come from OpenAI's own benchmarks unless noted; treat them as vendor-reported, not independent verification.
On Agents' Last Exam—professional computer-use tasks such as turning architectural drawings into a 3D model or logging an animated film shot by shot—GPT-6.1 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 roughly one-tenth its price per token. At max effort, OpenAI says Terra is over 3× as token-efficient as Claude Opus 5.5 and one-fourth its price per token, for about 9× lower cost per task (OpenAI's claim).
On MMMU-Pro, Terra beats GPT-5.6 Terra at every reasoning setting and nearly matches GPT-6 Astra and GPT-6.1 Sol. OSWorld 2.1 at medium effort surpasses GPT-5.6 Terra's best score at a fraction of the cost per task. For factuality, on OpenAI's internal error-inducing prompts benchmark, error rate at low effort fell from 53.6% (GPT-5.6 Terra) to 16.7%; at max effort it dropped 18 points to 4.3%.
Coding splits the lineup. On Terminal-Bench 4.0 at max effort, Terra reaches 50.3% versus GPT-5.6 Terra's 21.5%, but OpenAI still recommends GPT-6.1 Sol and GPT-6 Astra for the most complex coding tasks. Terminal-Bench-Science 0.1 at max reasoning surpasses GPT-6.1 Sol at lower cost and lands within five points of Opus 5.5 at under one-fifth the estimated cost.
Pricing and availability
API pricing for gpt-6.1-terra is $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens. Starting Oct 26, Plus, Pro, Enterprise, and Edu customers can use the model in ChatGPT Work and Codex.
AWS announced general availability on Amazon Bedrock the same day. Bedrock supports a context window up to 1 million input tokens, accepts text and image inputs, and allows ChatGPT Work and Codex to be configured to run Terra on Bedrock.
What it means for developers and teams
If your agents spend tokens on GUIs, documents, and mixed media, Terra is OpenAI's attempt to move the default away from the most expensive reasoning models. Halving API list prices versus GPT-5.6 Terra changes unit economics for high-volume automation without forcing a downgrade on vision-and-action tasks—at least on paper, per OpenAI's charts.
Bedrock availability matters for AWS-centric shops that need long context and a single inference endpoint. The 1M-token window plus image inputs fits ingestion-heavy pipelines (drawings, scans, long forms) that would otherwise need chunking.
For ChatGPT Work and Codex seats, the same allowance covering more tasks than GPT-5.6 Terra—and substantially more than GPT-6.1 Sol or GPT-6 Astra—pushes Terra as the daily driver while Sol and Astra stay the escalation path for hard code and max-reasoning jobs.
Caveats
OpenAI's cost and efficiency comparisons mix model tiers, reasoning effort levels, and token pricing; your workload may not see 9× savings. Agents' Last Exam and OSWorld are useful directional signals but do not guarantee production reliability on your stack. Factuality gains on internal prompts still leave double-digit error rates at low effort. For the hardest coding, OpenAI explicitly points buyers to GPT-6.1 Sol and GPT-6 Astra. Independent benchmarks and real-world red teaming should supplement launch-week numbers.