GPT-6.1 Terra is OpenAI’s new mid-tier model, released October 26, 2026, at $1 per million input tokens and $5 per million output tokens, half of GPT-5.6 Terra’s input price. Artificial Analysis measured its cost per task at $0.40 versus $1.40 for the old Terra. It now undercuts Claude Sonnet 5.5 by 50%, but every headline benchmark comes from OpenAI itself.
Key takeaways
- GPT-6.1 Terra lists at $1 input and $5 output per million tokens.
- Its cost per task fell to $0.40 from $1.40, per Artificial Analysis.
- It solved 50.3% of Terminal-Bench 4.0 tasks, up from 21.5%.
What is GPT-6.1 Terra?
GPT-6.1 Terra is the middle model in OpenAI’s four-tier lineup, sitting between the cheap Luna and the premium Sol and Astra. OpenAI built it for computer use: filling in forms, reading technical drawings, drafting proposals and running multi-step office tasks. It replaces GPT-5.6 Terra, which was left behind when the GPT-6 generation launched.
OpenAI calls it “a fast, efficient model that combines detailed visual reasoning and precise computer use,” according to 9to5Mac. The lineup runs Luna, Terra, Sol and Astra, from cheapest to most capable.
Terra was missing from last month’s GPT-6 release, when OpenAI shipped Sol and Luna. That left a gap in the middle of the price list. This launch fills it.
Who can use it today?
OpenAI said on X that GPT-6.1 Terra is rolling out to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex. Developers get it in the API as gpt-6.1-terra.
Free users are not on the list. According to Decrypt, the free tier stays on GPT-5.6 Luna. OpenRouter lists a 1.1 million token context window for the new model.
How much does GPT-6.1 Terra cost?
GPT-6.1 Terra costs $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens on OpenAI’s standard API tier. That is half of GPT-5.6 Terra’s $2 input price and 58% below its $12 output price. Against GPT-6 Astra, it is one-tenth the price per token.
OpenAI’s own line is that Terra comes “with 50% lower standard API prices.” The output cut is actually deeper. Artificial Analysis lists the old Terra at $2 input and $12 output, so output tokens fell from $12 to $5.
OpenRouter also shows two other tiers. Flex runs at $0.50 input and $2.50 output for slower jobs. Fast costs $2 and $10 for priority speed.
GPT-6.1 Terra vs rivals: API price per million tokens
The last column prices a sample monthly workload of 100 million input tokens and 20 million output tokens, before caching or batch discounts.
| Model | Input | Output | Sample month |
|---|---|---|---|
| GPT-6 Luna | $0.10 | $0.50 | $20 |
| GPT-6.1 Terra | $1.00 | $5.00 | $200 |
| GPT-5.6 Terra (old) | $2.00 | $12.00 | $440 |
| GPT-6.1 Sol | $2.00 | $10.00 | $400 |
| Claude Sonnet 5.5 | $2.00 | $10.00 | $400 |
| Gemini 4 Argon (intro) | $2.00 | $10.00 | $400 |
| Gemini 4 Argon (standard) | $4.00 | $20.00 | $800 |
| GPT-6 Astra | $10.00 | $50.00 | $2,000 |
The pattern is plain. At list price, Terra is now the cheapest serious mid-tier model from the three big US labs. It costs half of Sonnet 5.5 and half of Gemini 4 Argon’s introductory rate, which Google has said will rise to $4 and $20.
Is GPT-6.1 Terra better than GPT-5.6 Terra?
Yes, on every number OpenAI and Artificial Analysis have published. Terminal-Bench 4.0 coding scores more than doubled, factual errors fell sharply and cost per task dropped by about 71%. The catch: OpenAI still steers its hardest coding and research work to Sol and Astra, not Terra.
Here is what the published tests show:
- Coding: on Terminal-Bench 4.0 at maximum reasoning, Terra solved 50.3% of tasks versus 21.5% for GPT-5.6 Terra, OpenAI reported.
- Factual errors: on OpenAI’s internal error-eliciting test, the error rate at low effort fell from 53.6% to 16.7%. At maximum effort it fell to 4.3%.
- Professional tasks: on Agents’ Last Exam at medium reasoning, Terra scored 15 percentage points above the old Terra, per SBS.
- Hallucination: Artificial Analysis measured a 38% hallucination rate on AA-Omniscience, down from 88%.
- Speed: OpenAI says Terra finishes tasks more than three times faster on average than GPT-6 Astra and GPT-6.1 Sol.
How does it compare with GPT-6.1 Sol?
Sol is still ahead, but not by much. Artificial Analysis gave Terra 48 on its Coding Agent Index against 54 for Sol and 31 for GPT-6 Luna. On its broader Intelligence Index, Terra sits 4 points below Sol at maximum effort.
That is roughly 89% of Sol’s coding-agent score for half of Sol’s token price. For most business workloads, that trade is the whole story.
What does a task actually cost?
Per-token prices hide how many tokens a model burns. Artificial Analysis measured $0.40 per task at maximum effort, against $1.40 for GPT-5.6 Terra. Interesting Engineering, citing OpenAI, reported Terra’s task costs at max effort run roughly nine times lower than Claude Opus 5.5’s.
Should you trust the GPT-6.1 Terra benchmarks?
Treat them as a vendor pitch until more independent tests land. Most of the headline figures, including the error rates and the “three times faster” claim, come from OpenAI’s own evaluations. Artificial Analysis is the main outside check so far, and its numbers point the same way but are smaller in places.
Decrypt put it bluntly: “All of the figures are OpenAI’s own, and the error test is an internal one.” OpenAI has not published the task set behind the speed claim.
Two more caveats matter. First, the 50% cut applies to standard API prices, not to ChatGPT subscriptions. Second, OpenAI’s own guidance limits the model’s reach: “For in-depth research or complex coding, choose GPT-6.1 Sol or GPT-6 Astra,” the company wrote in its developer announcement.
The cybersecurity rating
Under OpenAI’s Preparedness Framework, Terra reached the “Critical” threshold for cybersecurity capability, the highest risk level, per SBS and Interesting Engineering. OpenAI says it carries the safeguards built for GPT-6 Astra.
That matters for money as well as safety. A cheap, fast model with top-tier hacking capability lowers the cost of attacks too. We have already covered how AI agents breached more than 100 firms at about $50 each. Cheaper capable models push that number down.
Who wins and who loses from GPT-6.1 Terra?
Buyers of AI agents win first: the same automation budget now buys roughly twice as many tokens. OpenAI gains a sharper weapon in its price war with Anthropic. The losers are rivals priced at $2 and $10, starting with Claude Sonnet 5.5, and anyone whose margins depend on reselling mid-tier tokens.
Winners
- Enterprise agent builders. Form-filling, document and back-office automation is exactly what Terra targets. Our sample workload falls from $440 to $200 a month.
- OpenAI’s sales team. OpenAI is already using discounts to win accounts. Replit moved a $100M+ deal away from Anthropic on price. A half-price Terra makes that pitch easier.
- Cloud resellers. Cheaper tokens usually mean more usage, which feeds the clouds that host OpenAI models.
Losers
- Anthropic. Claude Sonnet 5.5 lists at $2 and $10, double Terra. That pressure lands as Anthropic heads toward an IPO with a filing that shows $16.3B in revenue and an $80.7B loss.
- Google. Gemini 4 Argon’s introductory $2 and $10 price already matches Sonnet, not Terra. Its planned move to $4 and $20 looks harder to defend now.
- Cheap Chinese challengers. Models like Kimi K3.1 Pro sold on being far cheaper than US labs. That gap just narrowed.
The skeptical read on OpenAI itself: price cuts are not free. Each halving of list price needs usage to double just to hold revenue flat. A company whose valuation was just cut to $1.25 trillion is trading margin for share. That only pays off if the volume shows up.
What to watch next
No dated follow-up events have been announced for this launch. These are the concrete signals that will show whether GPT-6.1 Terra moves the market:
- Anthropic’s price response. A Sonnet price cut or new discount tier would confirm a mid-tier price war.
- Gemini 4 Argon’s standard pricing. Google said the $4 and $20 rate applies after the introductory period. Whether it holds after Terra is the test.
- Independent benchmarks. Watch for third-party scores on OSWorld and agent tasks that confirm or cut OpenAI’s numbers.
- GPT-5.6 Terra’s retirement. CloudPrice already lists the old model as deprecating. A shutdown date would force API customers to migrate.
FAQ
When was GPT-6.1 Terra released?
OpenAI announced GPT-6.1 Terra on October 26, 2026, and began rolling it out to ChatGPT and the API that day.
How much does the GPT-6.1 Terra API cost?
$1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens on the standard tier.
Is GPT-6.1 Terra free in ChatGPT?
No. It is available to Plus, Pro, Enterprise and Edu users. Free users stay on GPT-5.6 Luna, according to Decrypt.
Is GPT-6.1 Terra cheaper than Claude Sonnet 5.5?
Yes. At list price, Terra costs half as much: $1 and $5 per million tokens versus $2 and $10 for Sonnet 5.5.
Should I use Terra or GPT-6.1 Sol for coding?
OpenAI recommends Sol or Astra for complex coding. Terra scored 48 to Sol’s 54 on Artificial Analysis’ Coding Agent Index at half the price.
What is GPT-6.1 Terra’s context window?
OpenRouter lists a context window of 1.1 million tokens.
The bottom line
GPT-6.1 Terra is a price move dressed as a model launch. It delivers close to Sol-level agent performance at half of Sol’s price and cuts real per-task costs by about 71% versus its predecessor. For businesses running document and computer-use agents, it should be the new default to test.
Hold the benchmark claims loosely until outside tests catch up. The bigger signal is strategic: OpenAI is willing to undercut Anthropic and Google by half in the middle of the market. Expect the rest of the industry to answer on price.
Free guide · 12 pages
AI Tools Every Investor Should Use in 2026
10 tools for research, stock signals, charts and crypto, what each one costs, three ready-made stacks from $0 a month and 7 copy-paste prompts for filings and earnings calls. Enter your email and we will send it to you.
Plus one email a week: AI Money This Week, every Sunday. Unsubscribe anytime.
Sources
- OpenAI Developer Community: GPT-6.1 Terra announcement
- OpenAI on X: GPT-6.1 Terra rollout
- Artificial Analysis: GPT-6.1 Terra pushes the cost efficiency frontier
- OpenRouter: GPT-6.1 Terra pricing and specs
- 9to5Mac: OpenAI upgrades ChatGPT and Codex with GPT-6.1 Terra
- Decrypt: OpenAI launches GPT-6.1 Terra
- Interesting Engineering: GPT-6.1 Terra 50% cheaper
- SBS News: OpenAI unveils GPT-6.1 Terra
- CloudPrice: OpenAI Terra, Sol, Astra and Luna prices
- LangCopilot: Claude Sonnet 5.5 pricing
- eesel AI: Gemini 4 Argon pricing
Wealth Engine researches and drafts with AI tools and checks every figure against the sources above. How we report.
Leave a Reply