Compare
GPT-6.1 Sol vs Terra (2026): Which Model to Use in Codex
GPT-6.1 Terra landed in Codex and Copilot on Oct 26, 2026. Terra vs Sol on credits, messages per 5 hours, API price and which plans get it, with sources.
7 min read
Codex credits, message estimates and plan availability checked Oct 28, 2026 on Codex pricing; API prices on OpenAI API pricing; launch details on OpenAI’s Terra announcement; Copilot details on the GitHub changelog and Copilot models and pricing. Every source is listed at the bottom. No affiliate links on this page (how we work).
Quick answer
GPT-6.1 Terra costs exactly half of GPT-6.1 Sol in Codex credits and on the API, and OpenAI’s estimates work out to roughly twice the local messages per 5 hours on Plus (35–285 vs 15–160). Use Terra by default for everyday coding, terminal work and quick fixes. Switch to Sol when the task is genuinely hard, because OpenAI itself says “For the most complex coding tasks, GPT-6.1 Sol and GPT-6 Astra remain the better choice.”
Terra is in Codex on Plus, Pro, Business, Enterprise and Edu (the launch post lists Plus, Pro, Enterprise and Edu; the Codex pricing page adds Business). It is not on Free or Go, and it isn’t in regular ChatGPT Chat. In GitHub Copilot it’s on every paid plan, including the $10 Pro tier, which is more than Sol gets.
GPT-6.1 Sol vs Terra at a glance
| Model | Codex credits | API price | Plans |
|---|---|---|---|
| GPT-6.1 Terra | 25 in / 125 out | $1 in / $5 out | Plus, Pro, Business, Ent, Edu |
| GPT-6.1 Sol | 50 in / 250 out | $2 in / $10 out | Plus, Pro |
Codex credits are per 1M tokens (cached input: Terra 1.25, Sol 2.5). API prices are per 1M tokens, standard tier, short context up to 272K (cached input: Terra $0.05, Sol $0.10). Sources: Codex pricing and OpenAI API pricing, verified Oct 28, 2026. Sol’s “Plans” cell lists what the Plus card and the Pro Ultrafast table show; we didn’t confirm Sol plan-by-plan beyond that, so check your plan card.
Every line is a clean 50%: input, cached input, output, credits and dollars. That’s our arithmetic on OpenAI’s published numbers, and it matches OpenAI’s own line that Terra’s “standard API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol.”
On the API, long-context requests (over 272K) cost more for both: Terra $2 in / $7.50 out, Sol $4 in / $15 out per 1M. Batch is half again: Terra $0.50 / $2.50, Sol $1 / $5. Our OpenAI API pricing page has the full model list.
What Terra changes for Plus users
The headline for a $20 Plus user is message headroom. OpenAI’s Codex pricing page publishes these estimates for local messages per 5 hours on Plus (and Standard Business seats):
| Model | Messages / 5h |
|---|---|
| GPT-6 Luna | 350–3,000 |
| GPT-6.1 Terra | 35–285 |
| GPT-6.1 Sol | 15–160 |
| GPT-6 Sol | 15–150 |
| GPT-6 Astra | 5–45 |
Source: Codex pricing, verified Oct 28, 2026.
Three caveats from the same page, quoted because they matter:
- “These estimates are not fixed message limits”. Long tasks and big repos use more.
- “Weekly limits may also apply.” OpenAI doesn’t publish their size.
- “Pro plans currently have no five-hour limit.” So this table is mostly a Plus story.
The range is so wide that it’s less a promise and more a vibe. Michael Scott would call it “not superstitious, but a little stitious”: Terra will probably go further, and OpenAI won’t say how much further.
OpenAI’s launch post says it plainly: “In ChatGPT Work and Codex, your plan goes further with GPT-6.1 Terra—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.” The Codex page adds: “Overall Work and Codex plan limits are unchanged.” Translation: you didn’t get a bigger allowance. You got a cheaper model to spend it on.
When you do run out on Plus, you can buy credits, and OpenAI says “All users may also run extra local chats using an API key, with usage charged at standard API rates.” At the API rates above, Terra is the cheaper key to burn. Our Codex usage limits and Codex credits pages go deeper on the meters.
One small mess: on Oct 28 the Plus card on the Codex pricing page still lists “GPT-6.1 Sol and GPT-6 Luna”, while the feature-availability note on the same page says Terra is available on Plus. Trust the note and the launch post. The card just hasn’t caught up.
Not on Free or Go
If you’re on ChatGPT Free or Go ($8), Terra isn’t for you yet. Codex pricing lists both plans with “GPT-6 Luna at Standard speed in the desktop app, subject to rollout”, and Terra’s availability line names Plus, Pro, Business, Enterprise and Edu only.
Terra also “isn’t available in Chat”. It lives in ChatGPT Work and Codex. So even on Plus, you won’t find it in the normal chat model picker.
Joey doesn’t share food, and OpenAI doesn’t share Terra with the free tiers. Honestly, this one stings a bit: the model OpenAI pitches as its efficient one is exactly the model that would make sense on the cheap plans. For what Free actually includes, see Is Codex free?, and for the plan ladder, Codex pricing.
Two smaller footnotes from the Codex page: Terra “does not support UAE inference residency”, and the Ultrafast tier (Pro $500 only) lists GPT-6.1 Sol and GPT-6 Astra, not Terra.
Terra in GitHub Copilot: plans and multiplier
GitHub shipped Terra the same day. The changelog says it’s “generally available in GitHub Copilot” and “designed for everyday agentic coding and terminal work at a lower cost than GPT-6.1 Sol.”
Plans: “available to Copilot Pro, Pro+, Max, Business, and Enterprise users”. That’s a quiet win for the $10 Pro tier. In our Oct 24 check of GitHub’s supported-models docs, GPT-6.1 Sol was Pro+ and Max only (plus Business and Enterprise), not Pro. Terra is the way a Pro user gets a GPT-6.1 model.
Not on Free or Student: GitHub’s docs say “Copilot Student and Copilot Free users have access to models through auto model selection only”, and on Oct 28 the auto-selection list included GPT-6.1 Sol but not Terra. More in GitHub Copilot free limits.
Multiplier: there isn’t a premium-request multiplier on monthly plans anymore. Terra is “billed at provider list pricing under usage-based billing”, which means your AI credits drain at token rates (1 credit = $0.01):
| Model | Input / 1M | Output / 1M |
|---|---|---|
| GPT-6.1 Terra | $1.00 | $5.00 |
| GPT-6.1 Sol | $2.00 | $10.00 |
Up to 272K context. Cached input: Terra $0.05, Sol $0.10. Above 272K: Terra $2 / $7.50, Sol $4 / $15. Source: Copilot models and pricing, verified Oct 28, 2026.
In credits, that’s 100 credits per 1M input tokens and 500 per 1M output for Terra, double for Sol (our arithmetic). Annual Pro and Pro+ subscribers still on the older request-based billing have separate multipliers, which we didn’t check, so look at your own billing page if that’s you.
Rollout is gradual, and Business and Enterprise admins control it through the model policy. It shows up in VS Code, Visual Studio, Copilot CLI, cloud agent, the Copilot app, github.com, GitHub Mobile, JetBrains, Xcode and Eclipse. If you’re picking a Copilot tier around this, Copilot Pro vs Pro+ does the math.
When Sol is still the better pick
OpenAI’s own guidance is the useful bit here: “For the most complex coding tasks, GPT-6.1 Sol and GPT-6 Astra remain the better choice.” Terra is pitched as “a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing.”
OpenAI’s headline benchmark claim: Terra scores 50.3% on Terminal-Bench 4.0 at max effort, against 21.5% for GPT-5.6 Terra. Note that’s a comparison with the old Terra, not with Sol.
How I’d split it, using only what OpenAI and GitHub say each model is for:
- Terra: day-to-day edits, terminal work, quick agent tasks, anything where you’d rather have more messages than more brainpower.
- Sol (or Astra): gnarly multi-file refactors, hard debugging, the “most complex coding tasks” OpenAI itself names.
- On Plus: start in Terra and escalate. Burning Sol on a variable rename is GOB Bluth territory: you will look at your weekly meter and say “I’ve made a huge mistake.”
- On Pro: no five-hour limit, so the trade is mostly speed and weekly headroom. Terra still stretches the week further, per OpenAI.
For how Codex and Copilot stack up against everything else, see AI coding tools compared.
FAQ
Is GPT-6.1 Terra cheaper than Sol in Codex? Yes, Terra costs 25 credits per 1M input tokens and 125 per 1M output, exactly half of Sol’s 50 and 250 (checked Oct 28, 2026).
Is GPT-6.1 Terra on ChatGPT Plus? Yes, Terra is in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise and Edu, but not in regular Chat.
Is GPT-6.1 Terra in GitHub Copilot? Yes, it’s generally available on Copilot Pro, Pro+, Max, Business and Enterprise, billed at $1 input and $5 output per 1M tokens in AI credits.
Is GPT-6.1 Terra free? No, Terra isn’t on ChatGPT Free or Go, or on Copilot Free or Student, which use other models.
How many Terra messages do I get on Plus? OpenAI estimates 35–285 local messages per 5 hours on Plus, versus 15–160 for Sol, and says these aren’t fixed limits.
Sources
- OpenAI, Introducing GPT-6.1 Terra (availability, positioning, benchmark claim): https://openai.com/index/introducing-gpt-6-1-terra (verified 2026-10-28)
- OpenAI, Codex pricing (credits, messages per 5 hours, plan availability): https://developers.openai.com/codex/pricing (verified 2026-10-28)
- OpenAI, API pricing (gpt-6.1-terra and gpt-6.1-sol token prices): https://developers.openai.com/api/docs/pricing (verified 2026-10-28)
- GitHub Changelog, GPT-6.1 Terra in GitHub Copilot (plans, billing, surfaces): https://github.blog/changelog/2026-10-26-gpt-6-1-terra-in-github-copilot/ (verified 2026-10-28)
- GitHub Docs, Copilot models and pricing (per-token rates, credits): https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing (verified 2026-10-28)
- GitHub Docs, Supported AI models in Copilot (auto selection list, Free/Student access): https://docs.github.com/en/copilot/reference/ai-models/supported-models (verified 2026-10-28; Sol plan access verified 2026-10-24)