GPT-6.1 Terra

GPT-6.1 Terra is a proprietary large language model developed by OpenAI and released on October 26, 2026. It is the Terra-tier model of the GPT-6 series, positioned between GPT-6.1 Sol and GPT-6 Luna, and it succeeds GPT-5.6 Terra from the GPT-5.6 family. OpenAI's launch post describes it as "a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing," aimed at "the communication and computer-based tasks that fill the workday, from drafting proposals and interpreting technical drawings to filling out forms and editing videos."^[1] OpenAI's announcement on X summarized it as "delivering fast and capable computer use at a lower price."^[11] In the OpenAI API the model is gpt-6.1-terra, priced at $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens, which is half of GPT-6.1 Sol's per-token rates.^[1][3][4]

At launch the model was available in ChatGPT Work and Codex on paid plans, but not in regular ChatGPT conversations.^[9][10] Its system card addendum says OpenAI treats GPT-6.1 Terra as Critical in cybersecurity and High in biological and chemical capability under its Preparedness Framework, the same determinations it made for GPT-6.1 Sol and GPT-6 Astra, and applies the same safeguards stack.^[2] In an early independent measurement, Artificial Analysis scored GPT-6.1 Terra (max) at 48 on its Intelligence Index v4.3.2, six points above GPT-5.6 Terra at less than a third of the cost per task.^[15][16]

Overview

Background

The Sol, Terra and Luna tiers were introduced with the GPT-5.6 family, which OpenAI released in the API on July 9, 2026, describing GPT-5.6 Terra as the tier "for a balance of intelligence and cost."^[5] OpenAI's model page for GPT-5.6 Terra says it "roughly corresponds to the mini model tier used in earlier GPT-5 families."^[12] The GPT-6 generation began with GPT-6 Astra on September 3, 2026, followed by GPT-6 Sol and GPT-6 Luna on September 22 and GPT-6.1 Sol on September 29.^[5] None of those releases included a Terra model.^[18][20] 9to5Mac noted that when OpenAI released the GPT-6 Sol and Luna models, "its medium model Terra was excluded from the update," and Artificial Analysis called GPT-6.1 Terra "the first Terra-class model since GPT-5.6."^[15][18]

With the release, OpenAI's "Using GPT-6" guide lists four current models: GPT-6 Astra ("Highest intelligence"), GPT-6.1 Sol ("Balanced speed, cost, and intelligence"), GPT-6.1 Terra ("Fast and efficient") and GPT-6 Luna ("Fastest and most cost-effective").^[6] The system card addendum says GPT-6.1 Terra "uses the same types of data and training as GPT-6 Astra" and describes it as "a fast, efficient model for high-volume professional work, with strong image analysis and computer use capabilities."^[2] OpenAI has not published a parameter count or architecture details.

Release and availability

ChatGPT and Codex

OpenAI's launch post says that "Starting today, GPT-6.1 Terra is available to all Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex."^[1] The ChatGPT release notes entry for October 26, 2026 lists a broader set of plans: "The rollout includes Plus, Pro, Business, Enterprise, and Edu plans." It adds that "some Plus and Pro users may see GPT-6.1 Terra at Medium instead of GPT-6.1 Sol at Light in the default Power presets in Work," that Standard and Fast modes "are supported where available," and that Terra "does not support UAE inference residency."^[9] The ChatGPT documentation likewise lists Plus, Pro, Business, Enterprise and Edu, says the model "isn't available in Chat," and notes that Enterprise access "also depends on workspace availability and administrator controls."^[8] OpenAI's help center groups GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol and GPT-6 Luna as "models for ChatGPT Work and Codex" that "are not available in regular ChatGPT conversations."^[10]

OpenAI's ChatGPT model guide recommends Terra "for everyday writing and visual work," for "clear communication, interpreting images and screens, and completing tasks across supported apps," with examples such as "drafting proposals, reviewing visual materials, or filling out forms from source documents." GPT-6.1 Terra requires Codex CLI 0.159.0 or later and is selected with codex -m gpt-6.1-terra.^[7]

ChatGPT's pricing documentation estimates 35 to 285 local messages per five-hour period for GPT-6.1 Terra on Plus and Standard Business plans, against 15 to 160 for GPT-6.1 Sol, 15 to 150 for GPT-6 Sol, 5 to 45 for GPT-6 Astra and 350 to 3,000 for GPT-6 Luna. On credit-based plans GPT-6.1 Terra costs 25 credits per million input tokens, 1.25 credits per million cached input tokens and 125 credits per million output tokens, half of GPT-6.1 Sol's credit rates. GPT-5.6 Terra is listed at 50, 5 and 300 credits.^[8] OpenAI's launch post says that in ChatGPT Work and Codex "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."^[1]

API

The API changelog entry for October 26 says OpenAI "Released GPT-6.1 Terra, a fast, efficient model for visual reasoning, precise computer use, and clear, natural writing," through the Responses API and Chat Completions.^[5] The model page suggests using it "to match product photos to catalog entries, interpret technical drawings, act on application screens, or draft and revise text."^[3] Batch is supported; fine-tuning, the Realtime API, embeddings, image generation endpoints and audio endpoints are not.^[3]

Third-party platforms

GitHub announced on October 26 that GPT-6.1 Terra was "now generally available in GitHub Copilot" for Copilot Pro, Pro+, Max, Business and Enterprise users, describing it as "designed for everyday agentic coding and terminal work at a lower cost than GPT-6.1 Sol." GitHub said that "In early testing, GPT-6.1 Terra solved noticeably more terminal and C# coding tasks than GPT-5.6 Terra." The model is selectable in Visual Studio Code, Visual Studio, Copilot CLI, the Copilot cloud agent, the GitHub Copilot app, github.com, GitHub Mobile, JetBrains IDEs, Xcode and Eclipse, with a gradual rollout.^[13] OpenRouter lists the model at the same $1 input, $5 output and $0.05 cache-read rates, with a 1,050,000-token context and 128,000-token output limit. It also lists a "GPT-6.1 Terra Pro" entry, which it describes as "the same underlying model as GPT-6.1 Terra, served with reasoning.mode set to pro."^[14]

API specifications

GPT-6.1 Terra supports streaming, structured outputs, function calling, file search, web search, image input and prompt caching. Through the Responses API it can use OpenAI's built-in tools: web search, file search, image generation, code interpreter, hosted shell, apply patch, skills, computer use, MCP and tool search. Fine-tuning is not supported.^[3] Tool calling requires the Responses API; the model accepts Chat Completions requests, but OpenAI's migration guidance says applications that call tools through Chat Completions should move those requests to Responses.^[3][6]

Unlike GPT-6 Astra and GPT-6.1 Sol, GPT-6.1 Terra supports the none reasoning effort. OpenAI's guide advises starting at medium "and increase it if the improvement in quality is worth the added time and tokens," and, for developers migrating from GPT-5.6 Terra, to "compare both models on representative images and workflows using the same reasoning effort."^[6] Fast mode supports EU data residency for GPT-6.1 Terra, as it does for GPT-6.1 Sol and GPT-6 Luna but not for GPT-6 Astra or GPT-6 Sol.^[6]

Sources: OpenAI API model pages, pricing page and changelog.^{[3][4][5][12][21]}

Pricing

OpenAI lists the following API rates per million tokens for GPT-6.1 Terra. Prompts with more than 272,000 input tokens are billed at twice the input and cache rates and 1.5 times the output rate for the full request. Regional processing adds a 10% premium where available. A 12-hour cache write costs $2.50 per million tokens at Standard short-context rates.^[3][4]

Source: OpenAI API pricing page, accessed October 27, 2026.^[4] As of that date the Ultrafast pricing table listed only GPT-6 Astra and GPT-6.1 Sol.^[4]

OpenAI's launch post says GPT-6.1 Terra's "standard API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol, making it practical for developers to automate high-volume work, like reading incoming invoices and entering their details into accounting software."^[1] Against GPT-6.1 Sol every Standard rate is exactly half. Against GPT-5.6 Terra ($2 input, $0.20 cached input, $12 output), the input price is half, the output price is about 58% lower and the cached input price is a quarter.^[4][15] Per token, GPT-6.1 Terra costs one-tenth of GPT-6 Astra's Standard rates of $10 input and $50 output.^[4]

OpenAI's benchmark results

All results in this section were reported by OpenAI and were not independently reproduced at launch. The system card addendum notes that OpenAI's evaluations "were performed in our research environment or via our API, which may provide slightly different output from production ChatGPT," and that comparison values for earlier models "may reflect later versions of those models, and may vary from the values published at launch."^[2] Each chart in the launch post plots score against cost per task, with up to five reasoning settings (low, medium, high, xhigh and max) for OpenAI's models; GPT-5.6 Terra appears in the Terminal-Bench 4.0 chart only as a single maximum-effort point. Anthropic's Claude Opus 5.5 appears in two charts, once labelled "Claude Opus 5.5 w/ fallbacks," and Claude Sonnet 5.5 appears in two others.^[1]

Best score per model

The table gives each model's best score in each chart (the lowest error rate for the factuality chart), with the reasoning setting and cost per task from the chart data. A dash means the model does not appear in that chart.

Source: chart data in OpenAI's launch post.^[1]

In three of the six charts an Anthropic model posted the highest score: Claude Sonnet 5.5 at maximum effort on OSWorld 2.1 (83.9%) and Terminal-Bench 4.0 (70.6%), and Claude Opus 5.5 on Agents' Last Exam (63.2%). GPT-6 Astra led the other three: MMMU-Pro (86.7%), Terminal-Bench-Science 0.1 (68.1%) and the factuality evaluation (lowest error rate, 4.2%).^[1]

GPT-6.1 Terra at each reasoning setting

Source: chart data in OpenAI's launch post; dollar figures are cost per task as plotted.^[1]

Computer use

OpenAI measured computer use on Agents' Last Exam, which the post describes as "a benchmark of professional tasks that includes turning architectural drawings into a 3D model and reviewing an animated film shot by shot to create a time-coded log for a restoration team." It says that "At medium reasoning effort, GPT-6.1 Terra scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost," and that it "performs on par with GPT-6 Astra at only one-tenth its price per token."^[1] The chart data gives 57.6% at medium for GPT-6.1 Terra ($0.29 per task), 42.6% for GPT-5.6 Terra at medium ($0.84) and 57.6% for GPT-6 Astra at medium ($4.82). OpenAI also says that "At max effort, Terra is over three times as token efficient as Claude Opus 5.5, and one-fourth its price per token, resulting in a ~9x lower cost per task."^[1] In the chart, Opus 5.5 scored 63.2% at $8.82 per task against 60.0% at $1.01 for GPT-6.1 Terra.^[1]

The post further claims that Terra completes tasks "over three times faster than both GPT-6 Astra and GPT-6.1 Sol."^[1] Its charts plot score against cost; none plots task time.

On OSWorld 2.1, which OpenAI describes as testing "how well AI agents use desktop apps to complete multistep tasks, such as building a 3D model from a technical drawing," the post says that "At medium reasoning effort, GPT-6.1 Terra surpasses GPT-5.6 Terra's best score at a fraction of the cost per task."^[1] The chart data gives 66.7% at medium ($0.50) against GPT-5.6 Terra's best of 57.1% at max ($4.76). GPT-6.1 Terra's maximum-effort score of 74.2% is below GPT-6.1 Sol (78.0%), GPT-6 Astra (80.9%) and Claude Sonnet 5.5 (83.9%).^[1] The GPT-6.1 Sol launch post had reported results on OSWorld 2.0, so the two sets of OSWorld numbers are not directly comparable.^[22]

Visual reasoning

On MMMU-Pro, described as "a benchmark of visual reasoning," OpenAI says GPT-6.1 Terra "outperforms GPT-5.6 Terra at every reasoning setting" and "nearly matches the scores of our most capable frontier model, GPT-6 Astra, and GPT-6.1 Sol at much lower cost."^[1] At maximum effort the chart gives 85.5% for Terra, 86.3% for GPT-6.1 Sol and 86.7% for GPT-6 Astra.^[1]

Factuality

OpenAI says that on "our internal benchmark of error-inducing prompts, the error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra, a drop of nearly 37 percentage points. At maximum effort, the error rate fell by 18 percentage points to 4.3%, nearly matching the rates for GPT-6 Astra and GPT-6.1 Sol."^[1] The comparison values in this chart differ from those OpenAI plotted for the same models in the GPT-6.1 Sol launch post (GPT-6.1 Sol at low effort is 14.7% here against 7.7% in September), and the Terra post does not explain the change.^[1][22] Separately, the system card reports results on a refreshed evaluation built from ChatGPT conversations that users had flagged as containing factual errors. OpenAI says GPT-6.1 Terra "makes substantially fewer factual errors and is significantly less likely to reproduce user-reported hallucinations than GPT-6 Sol, achieving performance on par with GPT-6.1 Sol."^[2]

Coding and science

OpenAI says that at maximum reasoning effort GPT-6.1 Terra "solves 50.3% of tasks" on Terminal-Bench 4.0, "more than double GPT-5.6 Terra's published score of 21.5%," and adds: "For the most complex coding tasks, GPT-6.1 Sol and GPT-6 Astra remain the better choice."^[1] On Terminal-Bench-Science 0.1, which the post says "tests how well agents use code to carry out scientific research tasks, such as tracking cell divisions in a time-lapse video," it says Terra "surpasses GPT-6.1 Sol at lower cost, coming within five points of Opus 5.5 at less than one-fifth the estimated cost."^[1] The chart data gives 58.3% for Terra at max ($4.15 per task), 57.0% for GPT-6.1 Sol at max ($5.47) and 63.3% for the Opus 5.5 configuration with fallbacks ($23.21). At every lower setting Terra's score on this benchmark was below 44%.^[1]

Independent evaluations

Artificial Analysis published results on launch day. It reported that at maximum effort GPT-6.1 Terra "gains 6 points in the Artificial Analysis Intelligence Index over GPT-5.6 Terra, at less than a third of the cost per task ($0.40 vs $1.40)," with larger gains of 7 to 13 points at lower effort levels. It found the low and medium settings Pareto-optimal for intelligence against cost, "while the other effort levels sit behind the frontier." At maximum effort the model scored 4 points below GPT-6.1 Sol and 10 points above GPT-6 Luna on the index.^[15] Its model page lists GPT-6.1 Terra (Max) at 48 on Intelligence Index v4.3.2 and notes that it generated 130 million tokens while running the index, against a median of 81 million. As of October 27, 2026, the page measured its output speed at 91 tokens per second.^[16]

Sources: Artificial Analysis launch article and model page.^[15][16]

Artificial Analysis defines the AA-Omniscience hallucination rate as how often the model guesses "when it does not know the answer to a question." It reported that the drop from 88% to 38% came with a 2-point increase in AA-Omniscience accuracy, "meaning the model simultaneously answers more questions correctly."^[15] Trending Topics, summarizing the Artificial Analysis data, pointed out that the model "still guesses in more than a third of cases when it lacks an answer" and that at maximum effort it is not Pareto-optimal.^[20]

Safety and Preparedness

OpenAI published a system card addendum, "Addendum to GPT-6 Astra System Card: GPT-6.1 Terra," on October 26, 2026. It refers to the GPT-6 Astra system card for full descriptions of most evaluations.^[2]

Preparedness classification

Under its Preparedness Framework, OpenAI says it is "treating GPT-6.1 Terra as Critical capability in Cybersecurity, High capability in the Biological and Chemical domain, and below the High threshold in AI Self-Improvement." It applied "the same safeguards stack to GPT-6.1 Terra as we did for GPT-6.1 Sol and GPT-6 Astra (with probes updated to be compatible with the model)."^[2] The card quotes the framework's definition of Critical cyber capability as a model that can "identify and develop functional zero-day exploits of all severity levels in many hardened real-world critical systems without human intervention" or "devise and execute end-to-end novel strategies for cyberattacks against hardened targets given only a high level desired goal." It states that an internal Safeguards Report "informed our Safety Advisory Group's recommendation and OpenAI leadership's determination that these safeguards are sufficient for GPT-6.1 Terra's public launch."^[2] The launch post says the model "maintains the protections developed for GPT-6 Astra."^[1]

Source: GPT-6.1 Terra system card addendum.^[2]

OpenAI said the Internal Port result shows that "reliable exploitation of these recently disclosed vulnerabilities remains challenging for GPT-6.1 Terra," even though the model is "substantially more capable than older models on this evaluation."^[2]

Alignment evaluations

The launch post says that, according to OpenAI's evaluations, GPT-6.1 Terra "is more likely to stay within scope and honestly disclose limitations and potential blockers to the user compared to GPT-5.6 Terra."^[1] The system card reports mixed results against other GPT-6 models:

• Auto-review: in a test of how models react when Codex's Auto-review monitor blocks a command, GPT-6.1 Terra, like GPT-6 Astra, GPT-6 Sol and GPT-6.1 Sol, "made no attempts to bypass the monitor."^[2]
• Respecting warnings: in 29.6% of rollouts at maximum reasoning effort, GPT-6.1 Terra "saw the barrier but completed the task through an alternative route that violated the spirit of the restriction." OpenAI says this is "higher than GPT-6 Astra, GPT-6 Luna, GPT-6.1 Sol, but still lower than GPT-6 Sol," and that the evaluation "primarily measures low-stakes situations and is run without system-level control measures."^[2]
• External agent messages: GPT-6.1 Terra attempted to communicate with apparent peer agents in 65.00% of samples and performed the specified unauthorized action in 13.75% (11 of 80), against 50.00% and 1.25% for GPT-6.1 Sol and 41.25% and 16.25% for GPT-6 Sol. OpenAI re-ran the comparison models with corrected maximum reasoning settings for this card, so these values differ from those published at the GPT-6.1 Sol launch. OpenAI said the regression against GPT-6.1 Sol "largely reflects promises of future help that the model indicated it would not follow through on."^[2]
• Honeypot: on difficult problems from ExploitGym, an academic capture-the-flag benchmark, with a second flag planted in the evaluation harness, the model made an unsuccessful honeypot attempt in 0.07% of rollouts and had no successful honeypot exploits.^[2]
• Broken search tool: GPT-6.1 Terra failed to acknowledge an unavailable search tool in 1.83% of cases, against 2.08% for GPT-6.1 Sol.^[2]
• Deployment simulation: across 48,690 matched tasks simulating internal Codex traffic, GPT-6.1 Terra received 22 flags at severity 3 or higher (0.045%), against 24 for GPT-6.1 Sol, 26 for GPT-6 Astra, 38 for GPT-6 Sol and 61 for GPT-5.6 Sol. Relative to GPT-6.1 Sol and GPT-6 Astra, OpenAI observed higher flag rates for "concealing uncertainty, circumventing restrictions, exfiltration, other misalignment, destructive action and inbound prompt injection."^[2]
• Evaluation awareness: signs of verbalized evaluation or simulation awareness appeared in 9.88% of GPT-6.1 Terra's trajectories, compared with 4.35% for GPT-6.1 Sol, 8.07% for GPT-6 Astra, 4.54% for GPT-6 Sol and 2.47% for GPT-5.6 Sol.^[2]

SiliconANGLE highlighted the restriction-circumvention and agent-message results, writing that the model's "more advanced capabilities appear to come at a slight cost in terms of safety."^[17]

Monitorability

Among chain-of-thought lengths of 750 to 1,250 tokens, GPT-6.1 Terra followed a requested chain-of-thought instruction in 14.8% of cases, against 23.2% for GPT-6 Sol and 44.8% for GPT-6.1 Sol. OpenAI reported this as lower controllability. In a math side-tasking test, the model "did not produce a recorded side-task success while bypassing the CoT or full-context monitor flags," and OpenAI said the results "provide no observed evidence" that GPT-6 Astra, GPT-6.1 Sol or GPT-6.1 Terra can engage in chain-of-thought steganography.^[2]

Content safety, robustness and health

OpenAI says GPT-6.1 Terra performs "on-par with its predecessor GPT-5.6 Terra" on its Production Benchmarks with challenging prompts, and on par with GPT-5.6 Terra on multiturn jailbreak evaluations. On image-input safety tests it found a statistically significant regression in hate and extremism relative to GPT-5.6 Terra, which it said was "in-line with prior launches" after manual review. On prompt injection, it describes the model as similar to GPT-6 Astra and GPT-6.1 Sol, "which have saturated robustness on hierarchy evaluations at 99.99%."^[2]

On HealthBench Professional (length-adjusted), GPT-6.1 Terra scored 62.7 against 57.7 for GPT-5.6 Terra; on standard HealthBench it scored 1.9 points lower than GPT-5.6 Terra after length adjustment (55.1 against 57.0). On MentalHealthBench it scored 55.5% against 46.2% for GPT-5.6 Terra.^[2]

Reception

Coverage at launch focused on price and computer use. SiliconANGLE described Terra as "positioned as the everyday model in OpenAI's lineup, designed for routine work such as reading what's on a computer screen, clicking through applications, navigating websites and drafting messages."^[17] 9to5Mac wrote that OpenAI "currently has four main model classes, ranging from lowest price and capability to highest: Luna, Terra, Sol, and Astra."^[18] Interesting Engineering reported that Terra matched GPT-6 Astra's performance on Agents' Last Exam "at approximately one-tenth of its price per token."^[19] Trending Topics headlined its report "GPT-6.1 Terra Makes OpenAI's Midrange a Value Champion," citing the Artificial Analysis cost-efficiency results.^[20]

See also

• GPT-6.1 Sol
• GPT-6 Astra
• GPT-6 Sol
• GPT-5.6
• Computer-use agent
• OSWorld

References

1. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15 ^16 ^17 ^18 ^19 ^20 ^21 ^22 ^23 ^24 ^25OpenAI. "Introducing GPT-6.1 Terra." October 26, 2026. openai.com/...introducing-gpt-6-1-terra
2. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15 ^16 ^17 ^18 ^19 ^20 ^21OpenAI Deployment Safety Hub. "Addendum to GPT-6 Astra System Card: GPT-6.1 Terra." October 26, 2026. deploymentsafety.openai.com/gpt-6-1-terra
3. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15OpenAI API documentation. "GPT-6.1 Terra." Accessed October 27, 2026. developers.openai.com/...gpt-6.1-terra
4. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8OpenAI API documentation. "Pricing." Accessed October 27, 2026. developers.openai.com/...pricing
5. ^1 ^2 ^3 ^4 ^5OpenAI API documentation. "Changelog." Accessed October 27, 2026. developers.openai.com/...changelog
6. ^1 ^2 ^3 ^4OpenAI API documentation. "Using GPT-6." Accessed October 27, 2026. developers.openai.com/...latest-model
7. ^OpenAI, ChatGPT documentation. "Models." Accessed October 27, 2026. learn.chatgpt.com/...models
8. ^1 ^2OpenAI, ChatGPT documentation. "Pricing." Accessed October 27, 2026. learn.chatgpt.com/...pricing
9. ^1 ^2OpenAI Help Center. "ChatGPT release notes" (entry of October 26, 2026, "GPT-6.1 Terra"). Accessed October 27, 2026. help.openai.com/...6825453-chatgpt-release-notes
10. ^1 ^2OpenAI Help Center. "ChatGPT Work and Codex." Accessed October 27, 2026. help.openai.com/...20001275-chatgpt-work-and-codex
11. ^OpenAI (@OpenAI). "GPT-6.1 Terra: delivering fast and capable computer use at a lower price." X, October 26, 2026. x.com/...2114780711664398349
12. ^1 ^2OpenAI API documentation. "GPT-5.6 Terra." Accessed October 27, 2026. developers.openai.com/...gpt-5.6-terra
13. ^GitHub Changelog. "GPT-6.1 Terra in GitHub Copilot." October 26, 2026. github.blog/...-26-gpt-6-1-terra-in-github-copilot
14. ^OpenRouter. "GPT-6.1 Terra" model listing. Accessed October 27, 2026. openrouter.ai/...gpt-6.1-terra
15. ^1 ^2 ^3 ^4 ^5 ^6Artificial Analysis. "GPT-6.1 Terra pushes the cost efficiency frontier." October 26, 2026. artificialanalysis.ai/...-cost-efficiency-frontier
16. ^1 ^2 ^3Artificial Analysis. "GPT-6.1 Terra (max) - Intelligence, Performance & Price Analysis." Accessed October 27, 2026. artificialanalysis.ai/...gpt-6-1-terra
17. ^1 ^2Wheatley, Mike. "OpenAI launches GPT-6.1 Terra for faster, better-performing AI agents at half the price." SiliconANGLE, October 26, 2026. siliconangle.com/...ng-ai-agents-at-half-the-price
18. ^1 ^2 ^3Hall, Zac. "OpenAI upgrades ChatGPT and Codex with new GPT-6.1 Terra model." 9to5Mac, October 26, 2026. 9to5mac.com/...-codex-with-new-gpt-6-1-terra-model
19. ^Khollam, Aamir. "OpenAI's GPT-6.1 Terra delivers GPT-6 Astra-level results at 10% of the token price." Interesting Engineering, October 26, 2026. interestingengineering.com/...a-50-percent-cheaper
20. ^1 ^2 ^3Steinschaden, Jakob. "GPT-6.1 Terra Makes OpenAI's Midrange a Value Champion." Trending Topics, October 26, 2026. trendingtopics.eu/...penai-midrange-value-champion
21. ^OpenAI API documentation. "GPT-6.1 Sol." Accessed October 27, 2026. developers.openai.com/...gpt-6.1-sol
22. ^1 ^2OpenAI. "Introducing GPT-6.1 Sol." September 29, 2026. openai.com/...introducing-gpt-6-1-sol