Overview
GPT-6.1 Terra is the model OpenAI released on October 27, 2026, restoring the Terra tier to the 6.1 generation. Until then 6.1 had only Sol at $2 per million input tokens and $10 output; Terra brings those down to $1 and $5, cuts cached input to $0.05, and OpenAI puts it at roughly half the price of GPT-5.6 Terra.
The pitch is computer use. OpenAI says it beats GPT-6 Astra on vision tasks while running about 47 percent faster at roughly 92 percent lower cost. Artificial Analysis measured higher intelligence at under a third of the previous generation's cost, and Code Arena places the Terra Max variant 14th on its WebDev board with 1671 points.
Availability was immediate. The API and OpenRouter both carry it at the same rates, and ChatGPT Work and Codex 0.165.0 list GPT-6.1 Terra in their model catalogues. That Codex release also adds a Google Cloud Vertex AI provider and turns on memory v2 by default for new configurations.
Key Features
- Mid-tier pricing: $1 per million input tokens, $5 output, $0.05 for cached input. Both headline rates are half of GPT-6.1 Sol and half of GPT-5.6 Terra, while the cache rate undercuts both.
- Computer use as the focus: OpenAI aims this tier at reading screens, clicking controls and filling forms. The stated comparison is against Astra: better on vision, about 47 percent faster, about 92 percent cheaper per task.
- Third-party cost check: Artificial Analysis reports higher intelligence at under a third of the prior generation's cost. For products billed by token, that figure matters more than a benchmark score.
- Same-day surface area: The API and OpenRouter launched at identical pricing, and both ChatGPT Work and Codex 0.165.0 expose the model. Codex pairs it with a Vertex AI provider, decoupling model choice from deployment region.
- Codex side changes: Codex 0.165.0 enables memory v2 for new configurations, supports importing memory from external agents, and lets you change reasoning effort mid-conversation. The session picker gains repository filters and the agent command centre groups by directory.
Use Cases
- Screen-reading automations that click through interfaces, where per-task cost decides whether the workflow pays off.
- Batch processing of screenshots, receipts and UI states, the workload this tier is priced for.
- Repository chores in Codex, leaving harder reasoning to the higher tiers in the same family.
- A cheap execution slot in a model router, with hard prompts escalating to Astra.
- Token-billed SaaS with high cache hit rates, where the $0.05 cached input rate does the work.
- Teams deployed on Google Cloud, using Codex's Vertex AI provider.
Pros
- Input at $1 and output at $5, exactly half of GPT-6.1 Sol on both rates.
- Cached input falls to $0.05, which lowers the cost of long sessions with repeated context.
- OpenAI reports better vision than GPT-6 Astra at about 47 percent the latency and 92 percent less cost.
- Artificial Analysis puts cost under a third of the previous generation.
- Available on launch day through the API, OpenRouter, ChatGPT Work and Codex.
- The accompanying Codex release adds Vertex AI and memory v2 for both deployment and memory handling.
Pricing
The API charges $1 per million input tokens, $5 per million output tokens and $0.05 for cached input. Those are half of GPT-6.1 Sol at $2, $10 and $0.10, and roughly half of GPT-5.6 Terra at $2 and $12. OpenRouter lists the same $1 and $5 rates without markup. On the ChatGPT side the model ships inside existing plans, and Codex and ChatGPT Work users can pick it from the model catalogue. Quotas, rate limits and regional availability follow the official pricing page.
Summary
Terra closes the price gap in the 6.1 generation. Sol held the $2 and $10 line that most vendors converged on this quarter; Terra halves it, which changes the arithmetic for any workflow that reads a screen more than once per task. Showing up in Codex, ChatGPT Work and OpenRouter within a day signals OpenAI intends this as the default execution tier rather than a niche option. For anyone paying per token, it is the most concrete update of the past two days.
Version History
- GPT-6.1 Terra launches at $1 input and $5 output with $0.05 cached input (2026-10-27): OpenAI released GPT-6.1 Terra at $1 per million input tokens and $5 output, with cached input at $0.05, half the rates of GPT-6.1 Sol and roughly half of GPT-5.6 Terra. OpenAI reports better vision than GPT-6 Astra at about 47 percent the speed and 92 percent lower cost; Artificial Analysis puts cost under a third of the prior generation. Shipped the same day through the API, OpenRouter, ChatGPT Work and Codex, whose 0.165.0 release adds the model and a Google Cloud Vertex AI provider. Code Arena ranks Terra Max 14th on WebDev with 1671 points.