GPT-6.1 Terra vs GPT-6.1 Sol: which should you use?

What is GPT-6.1 Terra?

GPT-6.1 Terra is OpenAI's new mid-priced model, released Oct 26, 2026 (OpenAI's announcement on X). OpenAI's model page calls it "A fast, efficient model for visual reasoning, precise computer use, and clear, natural writing."

On price it sits between GPT-6.1 Sol and GPT-6 Luna. In the ChatGPT Learn guide, OpenAI pitches Terra "for everyday writing and visual work" and Sol "for repeated, long-running work" across code, apps and documents. Both take text and images, both have a 1,050,000-token context window, and Terra's knowledge cutoff is a month later (Jun 1, 2026 against Apr 30).

How much cheaper is GPT-6.1 Terra than Sol on the API?

Exactly half, on every line. Terra costs $1 input, $0.05 cached input and $5 output per 1M tokens, per OpenAI's API pricing. Sol costs $2, $0.10 and $10. Cache writes ($1.25 against $2.50 per 1M for 30-minute retention) and long prompts over 272K tokens follow the same 2-to-1 ratio.

Here's what that means over a month, using the same usage levels as our plan vs API calculator and 22 working days:

A coding assistant running all day comes to about $43 a month on Terra and $85 on Sol. Both models also get 50% off through Batch and Flex, so offline jobs cost half again.

Does GPT-6.1 Terra make Codex limits last longer?

Yes, roughly twice as long. OpenAI's Codex pricing page estimates 35 to 285 local Terra messages per 5 hours on Plus ($20) and Standard Business seats, against 15 to 160 for Sol. Terra also burns credits at half Sol's rate: 25 input, 1.25 cached and 125 output credits per 1M tokens, against 50, 2.5 and 250.

The plan itself didn't grow. OpenAI says "Overall Work and Codex plan limits are unchanged." Terra just spends the same allowance more slowly. Pro plans currently have no 5-hour limit, per the same page, though weekly limits may apply.

One trade-off: OpenAI lists Ultrafast mode for Sol and Astra in Codex, not Terra. Fast mode costs 2.5x included usage on any model that supports it.

Running low anyway? See what Codex credits cost and what to do when Codex hits its limit.

Is GPT-6.1 Terra as good as Sol for coding?

Nobody can say yet. One day after launch, Terra isn't on the LMArena leaderboards we track, while Sol is #4 of 23 on Code Arena among the models we price. We'll add Terra's ranks when it shows up.

OpenAI's own positioning is clear, though. Its Sol model page says to use Sol "for complex coding, computer use, and professional work", while Terra's pitch is about visuals, computer use and writing. For a big refactor or a long agent run, Sol is the safer pick. For drafting docs, reading screenshots or filling in forms from source files, OpenAI's guidance points to Terra, at half the cost.

Terra adds one extra setting: reasoning effort "none", for quick replies with no thinking step. Sol starts at "low".

How to use GPT-6.1 Terra in Codex and ChatGPT Work

• Codex CLI: update to version 0.159.0 or later, then run codex -m gpt-6.1-terra, or use /model inside a session.
• Default model: add model = "gpt-6.1-terra" to config.toml, which the CLI, IDE extension and desktop app share.
• ChatGPT desktop app and Work on the web: open Advanced under the composer and pick Terra from the model list.
• Plans: Plus, Pro, Business, Enterprise and Edu. It isn't on Free or Go, and it isn't in regular Chat.
• API: model ID gpt-6.1-terra. Tool calling needs the Responses API.

GitHub Copilot lists GPT-6.1 Terra on its Pro, Pro+ and Max plans, per its plans page.

Source: OpenAI's ChatGPT Learn models guide and Codex pricing page.

GPT-6.1 Terra or Sol: which should you pick?

• Pick Terra for writing, summaries of screenshots and charts, computer-use tasks, and everyday Codex work where you keep hitting the 5-hour cap on Plus.
• Pick Sol for hard coding: multi-file changes, debugging, long agent runs. It's OpenAI's pick for that work and the one with Ultrafast mode.
• Pick GPT-6 Luna for high-volume extraction and classification. It costs $0.10 in and $0.50 out per 1M tokens.

Our take: make Terra your default in Codex for anything that isn't serious coding, and switch to Sol when the task gets hard. On Plus, OpenAI's estimates suggest that roughly doubles how far a 5-hour window goes for the easy work. For Sol against OpenAI's top model, see GPT-6 Astra vs GPT-6.1 Sol.