GPT-6.1 Terra lowers the price of capable agent work
OpenAI’s October 26 release gives teams a cheaper model for coding and computer use, with independent tests supporting the cost case and safety evaluations still showing failures.
GPT-6.1 Terra is available in OpenAI’s API and in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu customers.
OpenAI positions it for routine coding, visual reasoning and computer use, reserving its larger models for the hardest work. Standard API prices are $1 per million input tokens and $5 per million output tokens; cached input costs $0.05 per million. [1][2]
Artificial Analysis found a substantial efficiency improvement in its October 26 tests: at maximum reasoning effort, Terra scored six points above GPT-5.6 Terra on the then-current Intelligence Index while costing about $0.40 per task, against $1.40 for its predecessor. That is an evaluator’s workload-specific result, not a promise about a customer’s bill. Its separate Coding Agent Index placed Terra below GPT-6.1 Sol. [3]
Take it further
Check Terra’s long-context and service-tier prices Public model documentation; API use requires an eligible account and is billed.
Sources & details 9 sources
Discuss in your voice app
Take the reporting, visual explanations and original sources into a conversation. Copy the brief, paste it into your preferred assistant, then start its voice mode.
Paste the brief into your voice app to discuss this reporting.Reporting sources
- OpenAI — Introducing GPT-6.1 Terra
Vendor release; availability and intended workloads.
- OpenAI API — GPT-6.1 Terra model card
Current price, context, tier and tool-calling conditions; checked October 28.
- Artificial Analysis — Terra cost-efficiency evaluation
October 26 independent evaluation; Intelligence Index v4.3.2, not v5.
- OpenAI — GPT-6.1 Terra system card
October 26 vendor safety evaluation; safeguard configuration limits interpretation.
- Artificial Analysis — Intelligence Index v5
October 27 methodology change; newer costs are not directly interchangeable.
- OpenAI API changelog — October 26 release
Dated release confirmation.
- Artificial Analysis — Intelligence Index methodology
Separate technical version history and task weighting; supports the v4.3.2/v5 boundary, not independent replication of historical costs.
- OpenAI API — Published pricing
Separate tariff record for base/cache-write rates, service tiers and regional processing; model-specific thresholds remain in the model card.
- OpenAI API — Computer-use execution controls
Implementation guidance requires application/runtime controls and confirmation of consequential actions; it does not reproduce the system-card failure rate.
Images & figures
Costs use AA’s October 26 v4.3.2 task mix at maximum reasoning effort. The later v5 workload differs.
Euangelion / Codex · Source · Usage terms · Download figure