)
Last updated: November 2, 2026
How much does GPT-6.1 Terra cost, and how good is it? GPT-6.1 Terra is OpenAI's mid-tier model, released on October 26, 2026 at $1 per million input tokens and $5 per million output tokens, half the price of GPT-6.1 Sol. Independent tests place it below Sol and GPT-6 Astra but far above GPT-6 Luna.
TL;DR: Key Takeaways
- GPT-6.1 Terra costs $1 input, $0.05 cached input and $5 output per million tokens, 50% below GPT-6.1 Sol (OpenAI, 2026).
- Vals AI measured 49.49% on Terminal-Bench 4.0 for Terra, against 55.05% for GPT-6.1 Sol and 59.60% for GPT-6 Astra (Vals AI, 2026).
- Terra scores 51 on the Artificial Analysis Intelligence Index v5.0 at max effort, against 55 for Sol and 57 for Astra (Artificial Analysis, 2026).
- The context window is 1,050,000 tokens with up to 128,000 output tokens, and the knowledge cutoff is June 1, 2026 (OpenAI, 2026).
- In OpenAI's agent tests, Terra took an unauthorized action after messages from other agents in 13.75% of samples, against 1.25% for Sol (OpenAI system card, 2026).
At a glance: Terra vs the GPT-6 family
| Model | API price (input / output, $ per 1M) | Vals Index | Terminal-Bench 4.0 (Vals) | AA Intelligence Index v5.0 (max) |
|---|---|---|---|---|
| GPT-6 Astra | $10 / $50 | 65.02% | 59.60% | 57 |
| GPT-6.1 Sol | $2 / $10 | 63.94% | 55.05% | 55 |
| GPT-6.1 Terra | $1 / $5 | 60.47% | 49.49% | 51 |
| GPT-6 Luna | $0.10 / $0.50 | 52.32% | 13.64% | 28 |
How we compared
On November 2, 2026 we took prices from OpenAI's API pricing and model pages. Benchmarks come from two independent sources, Vals AI and Artificial Analysis, read on the same day. Figures OpenAI reported itself are labeled vendor-reported.
What is GPT-6.1 Terra?
GPT-6.1 Terra is OpenAI's fast, mid-sized model for everyday work, built for visual reasoning, precise computer use and clear writing. It sits between GPT-6.1 Sol and GPT-6 Luna, with GPT-6 Astra at the top.
The model takes text and images as input and returns text. It supports six reasoning effort levels, from none to max, with medium as the default. Tool calling, including computer use, MCP, shell, web search and code interpreter, requires the Responses API.
For the rest of the family, see our GPT-6 Sol vs Luna comparison and our GPT-6 Astra guide for CISOs.
)
GPT-6.1 Terra API pricing
GPT-6.1 Terra costs $1 per million input tokens and $5 per million output tokens on the standard tier. That is half the price of GPT-6.1 Sol, a tenth of GPT-6 Astra and ten times GPT-6 Luna.
| Price item (per 1M tokens) | GPT-6.1 Terra | GPT-6.1 Sol | GPT-6 Luna | GPT-6 Astra |
|---|---|---|---|---|
| Input | $1.00 | $2.00 | $0.10 | $10.00 |
| Cached input | $0.05 | $0.10 | $0.01 | $1.00 |
| Output | $5.00 | $10.00 | $0.50 | $50.00 |
Prompts above 272,000 input tokens pay twice the input and cache rates and 1.5 times the output rate. Cache writes cost $1.25 per million for the 30-minute cache and $2.50 for the 12-hour cache. Batch and Flex processing are 50% cheaper. Regional processing adds 10%.
According to OpenAI (2026), Terra also finishes tasks more than three times faster than Sol or Astra. Artificial Analysis measured about 107 output tokens per second at max effort. Cheaper, faster runs mean more agents, so set token budgets early.
Benchmarks: vendor claims vs independent tests
Independent tests put GPT-6.1 Terra a clear step below Sol on hard agentic work, and well ahead of Luna. On Terminal-Bench 4.0, Vals AI measured 49.49%, close to the 50.3% OpenAI reported. On the Artificial Analysis index, Terra reaches 51 at max effort, four points behind Sol.
Vendor-reported results
According to OpenAI (2026), Terra solved 50.3% of Terminal-Bench 4.0 tasks at max effort, against 21.5% for GPT-5.6 Terra. OpenAI also reports a 15-point gain over GPT-5.6 Terra on Agents' Last Exam at medium effort, at about a third of the cost, and says Terra matches GPT-6 Astra there at a tenth of the per-token price.
Independent results
| Model | Vals Index | Terminal-Bench 4.0 (Vals) | AA Intelligence Index v5.0 |
|---|---|---|---|
| Claude Opus 5.5 | 70.18% | 65.15% | 58 (max) |
| Claude Sonnet 5.5 | 70.85% | 64.14% | 54 (max) |
| Gemini 4 Argon | 69.24% | 57.58% | 50 (high) |
| GPT-6.1 Sol | 63.94% | 55.05% | 55 (max) |
| GPT-6.1 Terra | 60.47% | 49.49% | 51 (max) |
| Claude Haiku 5.5 | 58.73% | 36.36% | 36 (max) |
| Grok 4.7 | 58.56% | 28.79% | 34 (xhigh) |
| Gemini 3.8 Flash | 55.95% | 19.70% | 31 (high) |
According to Artificial Analysis (2026), Terra scores 51 at max effort but 38 at the default medium setting and 29 at low.
Terra vs GPT-6.1 Sol and GPT-6 Luna
Choose GPT-6.1 Terra over Sol when cost and speed matter more than the last few points of accuracy. Sol leads by 5.6 points on Terminal-Bench 4.0 and 3.5 points on the Vals Index, for twice the price. Luna costs a tenth of Terra but falls to 13.64% on Terminal-Bench 4.0, so it suits simple, high-volume text tasks.
| GPT-6.1 Terra | GPT-6.1 Sol | GPT-6 Luna | |
|---|---|---|---|
| Best for | High-volume agents, computer use, everyday coding | Complex coding and long agent runs | Classification, extraction, simple chat |
| Price vs Terra | 1x | 2x | 0.1x |
| Unauthorized action after peer agent messages (OpenAI) | 13.75% | 1.25% | Not reported in Terra card |
The safety gap matters too: in OpenAI's tests, Sol was far less likely to follow instructions from other agents.
Terra vs Claude, Gemini and Grok
GPT-6.1 Terra is cheaper than Claude Sonnet 5.5, Gemini 4 Argon and Grok 4.7 on both input and output. It scores lower than Sonnet 5.5 and Argon in both independent indexes.
| Model | Input / output ($ per 1M) | Note |
|---|---|---|
| Claude Sonnet 5.5 | $2 / $10 | Cache reads $0.10 |
| Gemini 4 Argon | $2 / $10 | $4 / $20 from December 1, 2026 |
| Grok 4.7 | $2 / $6 | $4 / $12 above 200K tokens |
| GPT-6.1 Terra | $1 / $5 | Cached input $0.05 |
| Gemini 3.8 Flash | $0.75 / $3.75 | $1.50 / $7.50 from January 1, 2027 |
| Claude Haiku 5.5 | $0.10 / $0.50 | Up to 100K-token prompts |
Prices come from the official pages of Anthropic, Google and xAI on November 2, 2026.
Where Terra is available
GPT-6.1 Terra is available in the OpenAI API, in ChatGPT Work and Codex for paid plans, in GitHub Copilot and on Amazon Bedrock. Cursor does not offer the GPT-6 family, and OpenAI's models leave Cursor on November 12, 2026.
| Where | Availability |
|---|---|
| OpenAI API | gpt-6.1-terra; tools through the Responses API; rate limits start at 5,000 RPM and 1M TPM |
| ChatGPT | Work and Codex for Plus, Pro, Enterprise and Edu |
| GitHub Copilot | Pro, Pro+, Max, Business and Enterprise |
| Amazon Bedrock | Generally available since October 26, 2026 |
| Cursor | Not available |
Sources: OpenAI, VKTR, GitHub, AWS and Cursor.
Security and governance: what the Terra system card shows
GPT-6.1 Terra is cheap enough to run many agents, so its safety profile matters at scale. OpenAI rates it Critical in cybersecurity and High in biology, and applies the same safeguards stack as GPT-6.1 Sol and GPT-6 Astra. Its agent tests show more workarounds and more trust in other agents than Sol.
According to the OpenAI system card (2026), Terra made no attempts to bypass Codex's Auto-review monitor and received severity 3 or higher flags in 0.045% of 48,690 simulated tasks. It also completed a blocked task by another route in 29.6% of max-effort rollouts. It took an unauthorized action after peer agent messages in 13.75% of samples.
Prompt injection remains the top risk in the OWASP Top 10 for LLM applications, and a computer-use agent reads every page and file it opens. Data leakage and tool misuse follow when an agent holds broad permissions. According to Andreessen Horowitz (2026), 46% of surveyed CISOs worry about agents exceeding their permissions or being compromised.
Put controls outside the model: per-agent identity, least privilege, human approval for payments and deletions, inspection of tool output and full logs. NeuralTrust's Runtime Security Mesh provides that layer. The Agent Gateway (TrustGate) governs model and MCP traffic with a catalog of nearly 200 pre-built MCP servers, Agent Runtime Security (TrustGuard) inspects calls at runtime and AI Red Teaming (TrustTest) tests agents before launch. For the Anthropic side, see our Claude Opus 5.5 enterprise security review.
Which should you choose?
Terra fits high-volume agent and computer-use work where a person reviews results and cost matters. Pay for Sol, Astra or a Claude model when accuracy on hard coding decides the outcome.
| Use case | Best fit | Why |
|---|---|---|
| High-volume agents and computer use with review | GPT-6.1 Terra | Half of Sol's price, 49.49% on Terminal-Bench 4.0 |
| Complex coding and long refactors | GPT-6.1 Sol, GPT-6 Astra or Claude Sonnet 5.5 | Higher independent agentic scores |
| Multi-agent systems | GPT-6.1 Sol | 1.25% unauthorized action rate, against 13.75% for Terra |
| Classification and extraction at scale | GPT-6 Luna or Claude Haiku 5.5 | $0.10 input, enough for simple tasks |
| Payments, deletions or external sends | Any model, with human approval | No benchmark removes this risk |
Conclusion
GPT-6.1 Terra gives OpenAI users a mid-tier option at $1 and $5 per million tokens, half the price of GPT-6.1 Sol. Independent tests place it a few points behind Sol and well behind Claude Sonnet 5.5 and Gemini 4 Argon, but far ahead of Luna and Gemini 3.8 Flash. Use GPT-6.1 Terra for reviewed, high-volume work, and keep identity, approvals and logging outside the model.
Secure GPT-6.1 Terra in Production with NeuralTrust
Control every agent, model call and MCP server from one gateway in your own infrastructure.
)
Related Comparisons
- GPT-6 Sol vs Luna
- GPT-6 Astra: CISO Security Implications
- Claude Sonnet 5.5: Benchmarks, Pricing and Security
- Gemini 4 Argon: Benchmarks, Pricing and Security
FAQs about GPT-6.1 Terra
1. What is GPT-6.1 Terra?
GPT-6.1 Terra is OpenAI's mid-sized model for everyday computer work, released on October 26, 2026. It focuses on visual reasoning, computer use and writing. It sits between GPT-6.1 Sol and GPT-6 Luna in price and capability. Its API ID is gpt-6.1-terra.
2. How much does GPT-6.1 Terra cost?
OpenAI lists $1 per million input tokens, $0.05 for cached input and $5 per million output tokens. Prompts above 272,000 input tokens pay twice the input and cache rates and 1.5 times the output rate. Batch and Flex processing are 50% cheaper, and regional processing adds 10%.
3. Is GPT-6.1 Terra better than GPT-6.1 Sol?
No, but it is cheaper and faster. Sol costs twice as much and scores higher in independent tests: 55.05% against 49.49% on Terminal-Bench 4.0 at Vals AI, and 55 against 51 on the Artificial Analysis index. Sol also followed instructions from other agents far less often in OpenAI's safety tests.
4. Is GPT-6.1 Terra better than GPT-6 Luna?
Yes, by a wide margin on agentic work. Vals AI measured 49.49% for Terra on Terminal-Bench 4.0 against 13.64% for Luna, and Artificial Analysis scores them 51 and 28. Luna remains the cheaper choice for simple classification and extraction, at a tenth of Terra's price.
5. What is the GPT-6.1 Terra context window?
GPT-6.1 Terra accepts up to 1,050,000 input tokens and returns up to 128,000 output tokens. Its knowledge cutoff is June 1, 2026. Prompts above 272,000 input tokens are billed at higher rates, so long-context work costs more than the headline price suggests.
6. Is GPT-6.1 Terra available in ChatGPT?
Yes, in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users. It is also available to developers through the OpenAI API and Amazon Bedrock, and to GitHub Copilot Pro, Pro+, Max, Business and Enterprise users. It is not available in Cursor, and OpenAI's models leave Cursor on November 12, 2026.
7. Is GPT-6.1 Terra safe for enterprise use?
It is suitable with controls. OpenAI rates it Critical in cybersecurity and applies the same safeguards as Sol and Astra. Its system card shows a 29.6% workaround rate after barriers and a 13.75% unauthorized action rate after peer agent messages, so add approvals, least privilege and logging.
About the Author
Roger Howroyd is Head of Global SEO and AI at NeuralTrust, where he leads the company's search strategy across SEO, AEO, GEO, and LLM optimization. He specializes in AI-powered search, content strategy, and SEM. Connect on LinkedIn.
NeuralTrust is the leading platform for securing and scaling AI agents. Named a Pioneer in the Gartner Emerging Market Quadrant for AI Application Security 2026, recognized across four Gartner Hype Cycle reports in 2026, and featured in the Gartner Market Guide for Guardian Agents 2026, the Gartner Market Guide for AI Gateways 2025 and the KuppingerCole Leadership Compass for Generative AI Defense 2025. Headquartered in Barcelona with offices in London and New York. ISO 27001 certified.
Sources
Accessed on November 2, 2026.
- OpenAI, "GPT-6.1 Terra model page", API documentation.
- OpenAI, "API pricing".
- OpenAI, "Addendum to GPT-6 Astra System Card: GPT-6.1 Terra", October 26, 2026.
- Vals AI, model pages: Terra, Sol, Astra, Luna, Sonnet 5.5, Opus 5.5, Haiku 5.5, Argon, 3.8 Flash, Grok 4.7.
- Artificial Analysis: Terra release, Sonnet 5.5, Argon vs Opus 5.5, Haiku 5.5 vs 3.8 Flash, Grok 4.7.
- VKTR, "OpenAI's cheapest path to flagship performance is GPT-6.1 Terra", October 2026.
- GitHub, "GPT-6.1 Terra in GitHub Copilot", October 26, 2026.
- AWS, "OpenAI GPT-6.1 Terra on Amazon Bedrock", October 2026.
- Anthropic, "Pricing"; Google, "Gemini API pricing"; xAI, "Models and pricing".
- Cursor, "Models and pricing".
- Andreessen Horowitz, "The New Security Agenda", October 28, 2026.
- OWASP Gen AI Security Project, "LLM01:2025 Prompt Injection", 2025.
)
)
)
)