OpenAI: GPT-6.1 Terra
Fast, efficient model for visual reasoning, precise computer use, and clear, natural writing.
List input
$1
$/MTok · OpenAI
List output
$5
$/MTok
Context
1.1M
tokens
Sources
4
39 listings
Sources
Same model, different sources. Our first-party rate is listed first, then router and aggregator rows stay separate so you can see the spread instead of a single blended number.
| Source | Provider | Input | Output | Cache read | Context |
|---|---|---|---|---|---|
| tokenstatopenai/gpt-6.1-terra | OpenAI | $1 | $5 | $0.05 | 1.1M |
| OpenRoutercheapestopenai/gpt-6.1-terra:batch | OpenRouter | $0.5 | $2.50 | $0.025 | 1.1M |
| models.devopenai/gpt-6.1-terra | AI/ML API | $1 | $5 | $0.05 | 1.1M |
| models.devopenai/gpt-6.1-terra | Cloudflare | $1 | $5 | $0.05 | 1.1M |
| models.devopenai/gpt-6.1-terra | Eden AI | $1 | $5 | $0.05 | 1.1M |
| models.devopenai/gpt-6.1-terra | Kilo Gateway | $1 | $5 | $0.05 | 1.1M |
| models.devopenai/gpt-6.1-terra | LLM Gateway | $1 | $5 | $0.05 | 1.1M |
| models.devopenai/gpt-6.1-terra | Merge Gateway | $1 | $5 | $0.05 | 1.1M |
| models.devopenai/gpt-6.1-terra | NanoGPT | $1 | $5 | $0.05 | 1.1M |
| models.devopenai/gpt-6.1-terra | Ofox | $1 | $5 | $0.05 | 1.1M |
| models.devgpt-6.1-terra | OpenAI | $1 | $5 | $0.05 | 1.1M |
| OpenRouteropenai/gpt-6.1-terra | OpenRouter | $1 | $5 | $0.05 | 1.1M |
| models.devopenai/gpt-6.1-terra | Vercel AI Gateway | $1 | $5 | $0.05 | 1.1M |
| models.devglobal.openai.gpt-6.1-terra | Amazon Bedrock | $1 | $5 | $0.05 | 1.1M |
| models.devopenai-gpt-6.1-terra | DigitalOcean | $1 | $5 | $0.05 | 1.1M |
| models.devgpt-6.1-terra | Azure | $1 | $5 | $0.05 | 1.1M |
| models.devgpt-6.1-terra | Azure | $1 | $5 | $0.05 | 1.1M |
| models.devgpt-6.1-terra | CommandCode | $1 | $5 | $0.05 | 1.1M |
| models.devgpt-6.1-terra | GitHub Copilot | $1 | $5 | $0.05 | 1.1M |
| models.devgpt-6.1-terra | DevPass (LLM Gateway) | $1 | $5 | $0.05 | 1.1M |
| models.devgpt-6.1-terra | Requesty | $1 | $5 | $0.05 | 1.1M |
| models.devauto/openai/gpt-6.1-terra | LowRouter | $1 | $5 | $0.05 | 1.1M |
| LiteLLMglobal.openai.gpt-6.1-terra | N/A | $1 | $5 | $0.05 | N/A |
| LiteLLMopenai.gpt-6.1-terra | N/A | $1 | $5 | $0.05 | N/A |
| LiteLLMgpt-6.1-terra | N/A | $1 | $5 | $0.05 | N/A |
| LiteLLMazure_ai/gpt-6.1-terra | N/A | $1 | $5 | $0.05 | N/A |
| LiteLLMdatabricks-gpt-6-1-terra | N/A | $1 | $5 | $0.05 | N/A |
| LiteLLMdatabricks/databricks-gpt-6-1-terra | N/A | $1 | $5 | $0.05 | N/A |
| models.devopenai.gpt-6.1-terra | Amazon Bedrock | $1.10 | $5.50 | $0.055 | 1.1M |
| models.devus.openai.gpt-6.1-terra | Amazon Bedrock | $1.10 | $5.50 | $0.055 | 1.1M |
| models.devgpt-6.1-terra | Vivgrid | $1.10 | $5.50 | $0.055 | 1.1M |
| LiteLLMus.openai.gpt-6.1-terra | N/A | $1.10 | $5.50 | $0.055 | N/A |
| LiteLLMazure/us/gpt-6.1-terra | N/A | $1.10 | $5.50 | $0.055 | N/A |
| LiteLLMbedrock_mantle/openai.gpt-6.1-terra | N/A | $1.10 | $5.50 | $0.055 | N/A |
| LiteLLMus/gpt-6.1-terra | N/A | $1.10 | $5.50 | $0.055 | N/A |
| models.devgpt-6.1-terra | Cortecs | $1.18 | $5.88 | $0.059 | 1.1M |
| LiteLLMapac/gpt-6.1-terra | N/A | $1.20 | $6 | $0.06 | N/A |
| LiteLLMazure/apac/gpt-6.1-terra | N/A | $1.20 | $6 | $0.06 | N/A |
| LiteLLMeu/gpt-6.1-terra | N/A | $1.20 | $6 | $0.06 | N/A |
Capabilities
- Reasoning
- Tool calling
- Structured output
- Attachments
- In: text
- In: image
- In: pdf
- In: file
- Out: text
Knowledge cutoff: 2026-06-01Released: 2026-10-26Output limit: 128K tokensFamily: gpt-terra
Scores
Vendor launch report. Launch-chart results at max reasoning effort. Safety, health, alignment and preparedness values come from https://deploymentsafety.openai.com/gpt-6-1-terra (October 26, 2026); internal test rates are not production-traffic rates. Percentages, 0–1 safety scores, correlations and NLL retain their published scales. HealthBench adjusted/unadjusted, OSWorld offline, and ACE-normalized ExploitBench are distinct. Other launch efforts are separate benchmark rows.
Sparse coverage on purpose. Missing a score means we have not linked one yet, not that the model is unranked. Hover or tap the i next to any score for its source. Full boards on /benchmarks.
- Agents’ Last Exam v160i
- MMMU-Pro (Vision)85.5i
- OSWorld 2.1 (offline, OpenAI Oct 2026)74.2i
- Difficult-prompt factual error rate (%, lower is better)4.3i
- Terminal-Bench 4.050.3i
- Terminal-Bench Science 0.158.3i
- Side-task monitor bypass: CoT flag (conditional on success, %, lower is better)0i
- Side-task monitor bypass: action flag (conditional on success, %, lower is better)0i
- Side-task monitor bypass: full-context flag (conditional on success, %, lower is better)0i
- Production safe completion: Violent Illicit behavior (Oct 2026, 0–1)0.977i
- Production safe completion: Non-Violent Illicit behavior (Oct 2026, 0–1)0.997i
- Production safe completion: Extremism (Oct 2026, 0–1)0.936i
- Production safe completion: Hate (Oct 2026, 0–1)0.982i
- Production safe completion: Self-harm (standard) (Oct 2026, 0–1)0.974i
- Production safe completion: Gore (Oct 2026, 0–1)0.883i
- Production safe completion: Sexual (Oct 2026, 0–1)0.951i
- Production safe completion: Sexual/minors (Oct 2026, 0–1)0.965i
- Under-18 safety: Age-restricted goods, services, and dangerous challenges / activities (Oct 2026, 0–1)0.926i
- Under-18 safety: Sexual Content (Oct 2026, 0–1)0.973i
- Under-18 safety: Eating Disorders (Oct 2026, 0–1)0.938i
- Under-18 safety: Emotional Reliance (Oct 2026, 0–1)0.948i
- Under-18 safety: Self Harm (Oct 2026, 0–1)0.99i
- Under-18 safety: Gore (Oct 2026, 0–1)0.883i
- Agentic safe completion: Codex prod - Age-restricted actions (Oct 2026, 0–1)0.821i
- Agentic safe completion: Codex prod - Non-violent Wrongdoing (Oct 2026, 0–1)0.978i
- Agentic safe completion: Codex prod - Violent Wrongdoing (Oct 2026, 0–1)0.887i
- Agentic safe completion: Codex prod - Sensitive personal data (Oct 2026, 0–1)0.747i
- Agentic safe completion: Codex prod - Self-harm (Oct 2026, 0–1)0.92i
- Agentic safe completion: Chat prod - Chat Plugins (Oct 2026, 0–1)0.923i
- Agentic safe completion: Human red-teaming - Codex (Oct 2026, 0–1)0.958i
- Agentic safe completion: Human red-teaming - Chat Plugins (Oct 2026, 0–1)0.935i
- Vision safety: hate (Oct 2026, 0–1)0.985i
- Vision safety: extremism (Oct 2026, 0–1)0.928i
- Vision safety: self-harm (Oct 2026, 0–1)0.987i
- Vision safety: harms-erotic (Oct 2026, 0–1)0.991i
- Dynamic mental-health safety: Mental health (Oct 2026, 0–1)0.993i
- Dynamic mental-health safety: Emotional reliance (Oct 2026, 0–1)0.975i
- Dynamic mental-health safety: Self-harm (Oct 2026, 0–1)0.88i
- HealthBench overall (length-adjusted)55.1i
- HealthBench overall (unadjusted)51.1i
- HealthBench Hard (length-adjusted)31.4i
- HealthBench Hard (unadjusted)26i
- HealthBench Consensus (length-adjusted)95.1i
- HealthBench Consensus (unadjusted)94.8i
- MentalHealthBench (overall, %)55.5i
- MentalHealthBench (non-acute, %)56.5i
- MentalHealthBench (high acuity, %)56.5i
- MentalHealthBench (emergent, %)52.9i
- Indirect prompt-injection robustness (Oct 2026, %)98.77i
- Auto-review circumvention (Oct 2026, %, lower is better)0i
- Warning circumvention (Oct 2026, %, lower is better)29.6i
- ExploitGym honeypot: legitimate exploit (%)0.14i
- ExploitGym honeypot: unsuccessful attack (%, lower is better)0.07i
- ExploitGym honeypot: successful attack (%, lower is better)0i
- Blocker deception (Oct 2026, %, lower is better)8.47i
- Broken Search Tool failure (Oct 2026, %, lower is better)1.83i
- Unauthorized agent interaction: engagement (Oct 2026, %, lower is better)65i
- Unauthorized agent interaction: action (Oct 2026, %, lower is better)13.75i
- Unauthorized agent interaction: engagement without action (Oct 2026, %, lower is better)51.25i
- Realistic-work misaligned outcomes (Oct 2026, %, lower is better)2.28i
- Codex deployment simulation: severity 3+ (Oct 2026, %, lower is better)0.045i
- Codex simulation: severity 3+, no awareness (Oct 2026, %, lower is better)0.036i
- Codex simulation: evaluation awareness (Oct 2026, %)9.88i
- CoT controllability (750–1,250 tokens, %)14.8i
- Multimodal Troubleshooting Virology (pass@1, %)52.54i
- Multimodal Troubleshooting Virology (excluding refusals, %)46.52i
- AAV capsid packaging (Spearman correlation)0.5285i
- SHP2 protein function (global R²)0.3696i
- Coronavirus–ACE2 cell entry (composite score)0.4741i
- Phage–plasmid co-evolution (NLL, lower is better)10.8229i
- ExploitBench (ACE-normalized capability, %)97.1i
- ExploitBench internal port (Jun–Aug 2026, ACE success %)24.5i
- SEC-Bench Pro (pass@1, %)83.33i
- Sandbox-Bench (Oct 2026, success %)22.7i
- Sandbox-Bench runtime (Oct 2026, success %)60i
- Sandbox-Bench parser (Oct 2026, success %)14.3i
- Sandbox-Bench egress proxy (Oct 2026, success %)0i
- Internal Research Debugging (rubric score, %)76.39i
- KernelGen 1P (rubric score, %)45.34i
- Biology severe-risk safety (Oct 2026, 0–1)0.99i
- Biology dual-use safety (Oct 2026, 0–1)0.96i
- Biology benign non-overrefusal (Oct 2026, 0–1)0.982i
- Cyber safety: production Chat (Oct 2026, 0–1)0.98i
- Cyber safety: synthetic agentic environments (Oct 2026, 0–1)1i
- Cyber safety: semi-synthetic agentic environments (Oct 2026, 0–1)0.973i
- Realistic-work misalignment: security weakening (Oct 2026, %, lower is better)1.02i
- Realistic-work misalignment: data exfiltration (Oct 2026, %, lower is better)1.02i
- Realistic-work misalignment: destructive action (Oct 2026, %, lower is better)0.08i
- Realistic-work misalignment: unauthorized external communication (Oct 2026, %, lower is better)0.24i
- Realistic-work misalignment: service disruption (Oct 2026, %, lower is better)0i
- Realistic-work misalignment: unauthorized transactions (Oct 2026, %, lower is better)0i
- Realistic-work misalignment: circumventing restrictions (Oct 2026, %, lower is better)0i
- Codex simulation severity 1–2: ignoring instructions (Oct 2026, %, lower is better)11.803i
- Codex simulation severity 1–2: concealing uncertainty (Oct 2026, %, lower is better)0.429i
- Codex simulation severity 1–2: deception (Oct 2026, %, lower is better)0.06i
- Codex simulation severity 1–2: reward hacking (Oct 2026, %, lower is better)0.033i
- Codex simulation severity 1–2: circumvention (Oct 2026, %, lower is better)0.222i
- Codex simulation severity 1–2: exfiltration (Oct 2026, %, lower is better)0.142i
- Codex simulation severity 1–2: other (Oct 2026, %, lower is better)0.201i
- Codex simulation severity 1–2: destructive action (Oct 2026, %, lower is better)0.129i
- Codex simulation severity 1–2: credential harvesting (Oct 2026, %, lower is better)0.049i
- Codex simulation severity 1–2: sandbagging (Oct 2026, %, lower is better)0.018i
- Codex simulation severity 1–2: inbound prompt injection (Oct 2026, %, lower is better)0.021i
- Codex simulation severity 3+: circumvention (Oct 2026, %, lower is better)0.0246i
- Codex simulation severity 3+: reward hacking (Oct 2026, %, lower is better)0i
- Codex simulation severity 3+: exfiltration (Oct 2026, %, lower is better)0.0123i
- Codex simulation severity 3+: other (Oct 2026, %, lower is better)0.0082i
- Codex simulation severity 3+: destructive action (Oct 2026, %, lower is better)0i
- Codex simulation severity 3+: outbound prompt injection (Oct 2026, %, lower is better)0i
- Codex simulation severity 3+: credential harvesting (Oct 2026, %, lower is better)0i
- Codex simulation severity 3+: sabotage (Oct 2026, %, lower is better)0i
- HealthBench Professional64.2i
- HealthBench Professional (length-adjusted)62.7i
Arena rows are community-vote Elo from Arena AI, Artificial Analysis indices arrive via OpenRouter, and launch-eval figures are collected from vendor announcements, analyst writeups, and community benchmarks. Full boards and credits on /benchmarks.
See what this model costs you
The app reads the sessions your tools already wrote and prices this model's tokens at the published API rate. Free on your computer.