Comparison

GPT-6.1 Sol vs Terra: Which OpenAI Model Should You Use?

GPT-6.1 Sol vs Terra compared: Terra costs half as much ($1/$5 vs $2/$10), runs faster and wins some tests. Where Sol still leads, and which one to pick.

OpenAI now sells two GPT-6.1 models, and they sit right next to each other. GPT-6.1 Sol came out in late September. GPT-6.1 Terra arrived on October 26, 2026, at half the price.

So which one should you use? The short version: Terra for most everyday and computer-use work, Sol for the hardest coding and reasoning. Here is the longer version, with the numbers behind it.

TL;DR: GPT-6.1 Sol vs Terra comes down to cost against peak skill. Terra costs $1/$5 per million input/output tokens, half of Sol’s $2/$10, and OpenAI says it finishes computer tasks more than 3 times faster. On OpenAI’s Agents’ Last Exam chart, Terra even scores higher than Sol at each effort level. Sol is still OpenAI’s pick for the most complex coding, so use Terra by default and switch to Sol when a task is truly hard.

GPT-6.1 Sol vs Terra side by side

These specs come from OpenAI’s model pages for each model. Prices are per 1 million tokens at the standard API tier.

GPT-6.1 TerraGPT-6.1 Sol
TierMid (below Sol)Upper (below Astra)
ReleasedOctober 26, 2026September 29, 2026 (DevDay)
API namegpt-6.1-terragpt-6.1-sol
Input / cached / output$1 / $0.05 / $5$2 / $0.10 / $10
Context window1,050,000 tokens1,050,000 tokens
Max output128,000 tokens128,000 tokens
Knowledge cutoffJune 1, 2026April 30, 2026
Reasoning settingsnone, low, medium, high, xhigh, maxlow, medium, high, xhigh, max
Input typesText, imagesText, images
Built-in toolsSame list: computer use, MCP, web search, file search, code interpreter, hosted shell, apply patch, skills, image generation, tool search

Two details are easy to miss. Terra’s knowledge cutoff is about a month newer than Sol’s, so it knows a little more about recent events without searching. And Terra has a "none" reasoning setting, which skips thinking time for quick answers. Sol does not offer it.

GPT-6.1 Sol vs Terra price

Terra is exactly half of Sol on every line of the bill: input, cached input and output. Here is what that means for a simple job, like sending 100,000 tokens of documents in and getting 10,000 tokens of summary back:

  • Terra: $0.10 for input plus $0.05 for output, or about $0.15.
  • Sol: $0.20 for input plus $0.10 for output, or about $0.30.

The gap grows on agent tasks, because agents loop and use many tokens. OpenAI’s own cost-per-task estimates on Agents’ Last Exam, as read from its launch chart by CellCog, put Terra at $0.29 per task at medium effort against $1.31 for Sol. That is about 4.5 times cheaper per task, not just 2 times, because Terra also uses fewer tokens.

In ChatGPT, you do not pay per token, but the same idea applies. OpenAI says one usage allowance in ChatGPT Work and Codex covers "substantially more" tasks with Terra than with Sol.

Agentic AI, a team of expert AI agents on your iPhone. Skip the Sol vs Terra math and just describe the task you need done. Get the app free

GPT-6.1 Sol vs Terra benchmarks

Every number below is from OpenAI’s own testing. No independent lab has published a Sol vs Terra comparison yet.

Agents’ Last Exam (computer use)Terra scoreTerra cost/taskSol scoreSol cost/task
Low effort55.9%$0.1652.4%not stated
Medium effort57.6%$0.2953.8%$1.31
Max effort60.0%$1.0158.6%$2.00

On that test, Terra beats Sol at every effort level while costing far less. OpenAI also says Terra finishes these tasks more than 3 times faster than both Sol and GPT-6 Astra. For context, Claude Opus 5.5 scored 63.2% at max effort on the same chart, at an estimated $8.82 per task.

Other results OpenAI shared:

  • Terminal-Bench-Science 0.1: OpenAI says Terra beats Sol at a lower cost. It did not publish the exact scores.
  • MMMU-Pro (reading images and charts): Terra nearly matches Sol and Astra.
  • Wrong answers on tricky prompts: at max effort, Terra’s error rate is 4.3%, which OpenAI calls close to Sol and Astra.
  • Hard coding: Terra reached 50.3% on Terminal-Bench 4.0 at max effort, but OpenAI still says Sol and Astra are better for the most complex coding.

Which one should you use?

If your task is...PickWhy
Filling forms, clicking through apps, reading screenshotsTerraHigher computer-use score, faster, cheaper
Emails, proposals, summaries, everyday writingTerraOpenAI built it for "the workday," at half the price
Big agent jobs with many stepsTerraAbout 4.5x lower cost per task on OpenAI’s chart
Large, tricky codebases and long refactorsSolOpenAI still ranks Sol above Terra here
Legacy code migration like COBOLSolSol has public results, covered in our OpenAI COBOL post
Quick questions with no thinking neededTerraOnly Terra has the "none" reasoning setting

A simple rule works for most people: start with Terra, and move up to Sol only if Terra gets stuck. If you need the very top model, that is GPT-6 Astra, at $10/$50.

What this means for you

The interesting part of this launch is not that a cheaper model exists. It is that the cheaper model now wins on computer use, the skill agents need most. When a model that clicks, reads screens and fills forms gets faster and cheaper, agent apps can take on longer jobs for the same money.

It also means the "best" model changes from task to task. Knowing when to use Terra, Sol or Astra is real work, and most people should not have to do it.

Agentic AI is built around that idea. It gives you a team of ready-made expert agents on iPhone and iPad, up to 4 working on one job, and you just describe what you need. There is no account, and your first 3 tasks are free. Learn the basics in what is agentic AI.

Agentic AI, one request, a team of agents, no model picking. Try your first tasks free. Get the app free

Quick answers

Is GPT-6.1 Terra better than GPT-6.1 Sol? For computer use, yes: OpenAI’s Agents’ Last Exam chart shows Terra ahead at every effort level. For the hardest coding, OpenAI still says Sol is better.

Is GPT-6.1 Terra cheaper than Sol? Yes, by half. Terra is $1 in and $5 out per million tokens. Sol is $2 in and $10 out.

GPT-5.6 Terra vs GPT-6.1 Sol: which is better? GPT-6.1 Sol. It is also cheaper than GPT-5.6 Terra ($2/$10 vs $2/$12). Both are now beaten on price by GPT-6.1 Terra.

Do GPT-6.1 Sol and Terra have the same context window? Yes. Both take up to 1,050,000 tokens and can write up to 128,000 tokens back.

Which is faster, Sol or Terra? Terra. OpenAI says it completes computer-use tasks more than 3 times faster than Sol.

Can I use both in ChatGPT? Paid plans can. GPT-6.1 Terra is in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users. It is not on the Free plan.


Sources: OpenAI: Introducing GPT-6.1 Terra · OpenAI model page: GPT-6.1 Terra · OpenAI model page: GPT-6.1 Sol · OpenAI model page: GPT-5.6 Terra · 9to5Mac · iThinkDifferent · CellCog (chart data)

Stop working alone.

Give any task to a team of AI experts on your iPhone. Start with 3 free tasks, then Agentic AI PRO from $34.99 a year.

Download free on iPhone On your computer?Point your iPhone camera here to install.

3 free tasks · No account · No login