# Fortress > Source: Benchize (https://benchize.com) — https://benchize.com/benchmarks/scale%3Afortress. When citing or reusing this content, credit "Benchize" and link to https://benchize.com/benchmarks/scale%3Afortress. The rapid advancement of large language models (LLMs) introduces powerful dual-use capabilities that could both threaten and bolster national security and public safety (NSPS). Developers often implementmodel safeguardsto help protect against misuse of models that could lead to possible risks. However, these mitigation measures also sometimes inadvertently prevent models from providing useful information. We need to understand the extent to which model safeguards bothprevent harmful responses and enable helpful onesto assess the relevant trade-offs for research, development, and policy. Existing benchmarks often do not adequately test model robustness to national security and public safety (NSPS) related risks in a scalable, objective manner that accounts for the dual-use nature of NSPS information. To address this, Scale AI introduces FORTRESS (Frontier Risk Evaluation for National Security and Public Safety), a benchmark featuring over 1,010 expert-crafted adversarial prompts designed to evaluate the safeguards of frontier LLMs (500 in the public dataset). FORTRESS assesses model responses across three domains: Chemical, Biological, Radiological, Nuclear and Explosive (CBRNE); Political Violence & Terrorism; and Criminal & Financial Illicit Activities. Canonical URL: https://benchize.com/benchmarks/scale%3Afortress Last updated: Wed, 21 Oct 2026 18:52:20 GMT Reproduction and quotation must name Benchize, link to the canonical HTML URL, and retain individual source citations. The rapid advancement of large language models (LLMs) introduces powerful dual-use capabilities that could both threaten and bolster national security and public safety (NSPS). Developers often implementmodel safeguardsto help protect against misuse of models that could lead to possible risks. However, these mitigation measures also sometimes inadvertently prevent models from providing useful information. We need to understand the extent to which model safeguards bothprevent harmful responses and enable helpful onesto assess the relevant trade-offs for research, development, and policy. Existing benchmarks often do not adequately test model robustness to national security and public safety (NSPS) related risks in a scalable, objective manner that accounts for the dual-use nature of NSPS information. To address this, Scale AI introduces FORTRESS (Frontier Risk Evaluation for National Security and Public Safety), a benchmark featuring over 1,010 expert-crafted adversarial prompts designed to evaluate the safeguards of frontier LLMs (500 in the public dataset). FORTRESS assesses model responses across three domains: Chemical, Biological, Radiological, Nuclear and Explosive (CBRNE); Political Violence & Terrorism; and Criminal & Financial Illicit Activities. Safety & robustness · Models tested: 31 · Score publishers: [scale](https://labs.scale.com/leaderboard/fortress) ## Model results | # | Model | Score | ± | | | --- | --- | --- | --- | --- | | 1 | [DeepSeek R1](https://benchize.com/models/deepseek-r1.md) | 74.39source score | CI 71.67–77.11 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 74.39 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration DeepSeek R1 sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 71.67–77.11 contaminationWarning ORS score: 0.49 rank 71 reportedConfidenceHalfWidth 2.72 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label DeepSeek R1 Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/70` | | | | | | 2 | [Qwen3 235B A22B](https://benchize.com/models/qwen3-235b-a22b.md) | 61.39source score | CI 58.03–64.75 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 61.39 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration Qwen3-235B-A22B sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 58.03–64.75 contaminationWarning ORS score: 0.98 rank 68 reportedConfidenceHalfWidth 3.36 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label Qwen3-235B-A22B Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/67` | | | | | | 3 | [DeepSeek V3.1](https://benchize.com/models/deepseek-v3-1.md) | 60.55source score | CI 57.19–63.91 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 60.55 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration DeepSeek-V3.1 sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 57.19–63.91 contaminationWarning ORS score: 1.76 rank 67 reportedConfidenceHalfWidth 3.36 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label DeepSeek-V3.1 Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/66` | | | | | | 4 | [GPT-4.1 Mini](https://benchize.com/models/gpt-4-1-mini.md) | 59.19source score | CI 55.89–62.49 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 59.19 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration GPT-4.1 mini sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 55.89–62.49 contaminationWarning ORS score: 1.38 rank 65 reportedConfidenceHalfWidth 3.3 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label GPT-4.1 mini Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/64` | | | | | | 5 | [Gemini 1.5 Pro](https://benchize.com/models/gemini-1-5-pro.md) | 53.86source score | CI 50.64–57.08 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 53.86 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration Gemini 1.5 Pro sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 50.64–57.08 contaminationWarning ORS score: 3.17 rank 60 reportedConfidenceHalfWidth 3.22 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label Gemini 1.5 Pro Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/59` | | | | | | 6 | [GPT-4.1](https://benchize.com/models/gpt-4-1.md) | 53.02source score | CI 49.51–56.53 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 53.02 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration GPT-4.1 sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 49.51–56.53 contaminationWarning ORS score: 2.87 rank 59 reportedConfidenceHalfWidth 3.51 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label GPT-4.1 Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/58` | | | | | | 7 | [Gemini 3.1 Flash Lite Preview](https://benchize.com/models/gemini-3-1-flash-lite-preview.md) | 51.05source score | CI 47.57–54.53 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 51.0542 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration gemini-3.1-flash-lite-preview sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 47.57–54.53 contaminationWarning ORS score: 2.15 rank 58 reportedConfidenceHalfWidth 3.48 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label gemini-3.1-flash-lite-preview Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/57` | | | | | | 8 | [Gemini 1.5 Flash](https://benchize.com/models/gemini-1-5-flash.md) | 50.61source score | CI 47.33–53.89 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 50.61 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration Gemini 1.5 Flash sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 47.33–53.89 contaminationWarning ORS score: 4.45 rank 57 reportedConfidenceHalfWidth 3.28 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label Gemini 1.5 Flash Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/56` | | | | | | 9 | [GPT-4o Mini](https://benchize.com/models/gpt-4o-mini.md) | 48.07source score | CI 44.76–51.38 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 48.07 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration GPT-4o mini sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 44.76–51.38 contaminationWarning ORS score: 2.27 rank 56 reportedConfidenceHalfWidth 3.31 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label GPT-4o mini Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/55` | | | | | | 10 | [GPT-4o](https://benchize.com/models/gpt-4o.md) | 47.18source score | CI 43.82–50.54 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 47.18 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration GPT-4o sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 43.82–50.54 contaminationWarning ORS score: 1.68 rank 55 reportedConfidenceHalfWidth 3.36 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label GPT-4o Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/54` | | | | | | 11 | [Llama 4 Maverick](https://benchize.com/models/llama-4-maverick.md) | 40.09source score | CI 36.73–43.45 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 40.09 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration Llama 4 Maverick sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 36.73–43.45 contaminationWarning ORS score: 4.95 rank 50 reportedConfidenceHalfWidth 3.36 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label Llama 4 Maverick Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/49` | | | | | | 12 | [Claude 3.7 Sonnet](https://benchize.com/models/claude-3-7-sonnet.md) | 38.01source score | CI 34.61–41.41 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 38.01 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration Claude 3.7 Sonnet sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 34.61–41.41 contaminationWarning ORS score: 4.45 rank 49 reportedConfidenceHalfWidth 3.4 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label Claude 3.7 Sonnet Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/48` | | | | | | 13 | [Claude 3.5 Haiku](https://benchize.com/models/claude-3-5-haiku.md) | 30.41source score | CI 27.21–33.61 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 30.41 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration Claude 3.5 Haiku sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 27.21–33.61 contaminationWarning ORS score: 13.16 rank 48 reportedConfidenceHalfWidth 3.2 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label Claude 3.5 Haiku Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/47` | | | | | | 14 | [O3 Mini](https://benchize.com/models/o3-mini.md) | 30.05source score | CI 26.75–33.35 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 30.05 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration o3-mini sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 26.75–33.35 contaminationWarning ORS score: 5.65 rank 46 reportedConfidenceHalfWidth 3.3 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label o3-mini Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/45` | | | | | | 15 | [Gemini 3.1 Pro Preview](https://benchize.com/models/gemini-3-1-pro-preview.md) | 29.76source score | CI 26.47–33.05 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 29.7642 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration gemini-3.1-pro-preview sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 26.47–33.05 contaminationWarning ORS score: 3.13 rank 45 reportedConfidenceHalfWidth 3.29 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label gemini-3.1-pro-preview Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/44` | | | | | | 16 | [Claude Opus 4](https://benchize.com/models/claude-opus-4.md) | 27.61source score | CI 24.39–30.83 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 27.61 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration Claude Opus 4 sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 24.39–30.83 contaminationWarning ORS score: 2.52 rank 43 reportedConfidenceHalfWidth 3.22 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label Claude Opus 4 Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/42` | | | | | | 17 | [Claude Sonnet 4](https://benchize.com/models/claude-sonnet-4.md) | 24.37source score | CI 21.3–27.44 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 24.37 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration Claude Sonnet 4 sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 21.3–27.44 contaminationWarning ORS score: 5.04 rank 36 reportedConfidenceHalfWidth 3.07 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label Claude Sonnet 4 Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/35` | | | | | | 18 | [O4 Mini](https://benchize.com/models/o4-mini.md) | 21.49source score | CI 18.43–24.55 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 21.49 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration o4-mini sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 18.43–24.55 contaminationWarning ORS score: 5.14 rank 33 reportedConfidenceHalfWidth 3.06 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label o4-mini Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/32` | | | | | | 19 | [Claude Opus 4.6](https://benchize.com/models/claude-opus-4-6-off.md)Non-thinking | 20.52source score | CI 17.55–23.49 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 20.52 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort non-thinking Model configuration claude-opus-4-6 (Non-Thinking) sourceModelVersion Not reported thinkingEnabled false ### Reported statistics Uncertainty Confidence interval: 17.55–23.49 contaminationWarning Not reported rank 31 reportedConfidenceHalfWidth 2.97 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label claude-opus-4-6 (Non-Thinking) Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/30` | | | | | | 20 | [GPT-5.6 Sol](https://benchize.com/models/gpt-5-6-sol.md) | 20.14source score | CI 18.32–21.96 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 20.14 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration gpt-5.6-sol sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 18.32–21.96 contaminationWarning ORS score: 5.34 rank 29 reportedConfidenceHalfWidth 1.82 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label gpt-5.6-sol Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/28` | | | | | | 21 | [O1](https://benchize.com/models/o1.md) | 19.38source score | CI 16.53–22.23 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 19.38 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration o1 sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 16.53–22.23 contaminationWarning ORS score: 5.05 rank 28 reportedConfidenceHalfWidth 2.85 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label o1 Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/27` | | | | | | 22 | [gpt-oss-20b](https://benchize.com/models/gpt-oss-20b.md) | 17.62source score | CI 14.83–20.41 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 17.62 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration gpt-oss-20b sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 14.83–20.41 contaminationWarning ORS score: 11.94 rank 25 reportedConfidenceHalfWidth 2.79 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label gpt-oss-20b Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/24` | | | | | | 23 | [Claude Opus 5](https://benchize.com/models/claude-opus-5.md) | 16.84source score | CI 14.91–18.77 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 16.84 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration claude-opus-5 sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 14.91–18.77 contaminationWarning ORS Score: 7.91 rank 20 reportedConfidenceHalfWidth 1.93 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label claude-opus-5 Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/19` | | | | | | 24 | [O3](https://benchize.com/models/o3.md) | 16.01source score | CI 13.33–18.69 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 16.01 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration o3 sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 13.33–18.69 contaminationWarning ORS score: 7.42 rank 17 reportedConfidenceHalfWidth 2.68 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label o3 Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/16` | | | | | | 25 | [GPT-6 Sol](https://benchize.com/models/gpt-6-sol.md) | 13.46source score | CI 12.06–14.86 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 13.46 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration gpt-6-sol sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 12.06–14.86 contaminationWarning ORS Score: 4.06 rank 11 reportedConfidenceHalfWidth 1.4 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label gpt-6-sol Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/10` | | | | | | 26 | [Claude Opus 5.5](https://benchize.com/models/claude-opus-5-5.md) | 13.18source score | CI 11.41–14.95 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 13.18 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration claude-opus-5-5 sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 11.41–14.95 contaminationWarning ORS Score: 9.50 rank 10 reportedConfidenceHalfWidth 1.77 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label claude-opus-5-5 Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/9` | | | | | | 27 | [Claude 3.5 Sonnet](https://benchize.com/models/claude-3-5-sonnet.md) | 12.96source score | CI 10.62–15.3 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 12.96 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration Claude 3.5 Sonnet sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 10.62–15.3 contaminationWarning ORS score: 21.07 rank 8 reportedConfidenceHalfWidth 2.34 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label Claude 3.5 Sonnet Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/7` | | | | | | 28 | [Muse Spark 1.1](https://benchize.com/models/muse-spark-1-1.md) | 12.4source score | CI 10.92–13.88 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 12.4 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration Muse Spark 1.1 sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 10.92–13.88 contaminationWarning ORS score: 6.92 rank 6 reportedConfidenceHalfWidth 1.48 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label Muse Spark 1.1 Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/5` | | | | | | 29 | [GPT-6 Luna](https://benchize.com/models/gpt-6-luna.md) | 10.78source score | CI 9.56–12 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 10.78 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration gpt-6-luna sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 9.56–12 contaminationWarning ORS score: 7.62 rank 5 reportedConfidenceHalfWidth 1.22 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label gpt-6-luna Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/4` | | | | | | 30 | [Claude Haiku 5.5](https://benchize.com/models/claude-haiku-5-5.md) | 9.48source score | CI 8.14–10.82 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 9.48 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration Claude Haiku 5.5 sourceModelVersion claude-haiku-5-5 thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 8.14–10.82 contaminationWarning ORS score: 9.00 rank 3 reportedConfidenceHalfWidth 1.34 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label Claude Haiku 5.5 Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/2` | | | | | | 31 | [gpt-oss-120b](https://benchize.com/models/gpt-oss-120b.md) | 8.24source score | CI 6.31–10.17 | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 8.24 \| source_score \| reported \| Higher is better \| ### Configuration Reasoning effort unknown Model configuration gpt-oss-120b sourceModelVersion Not reported thinkingEnabled Not reported ### Reported statistics Uncertainty Confidence interval: 6.31–10.17 contaminationWarning ORS score: 18.00 rank 2 reportedConfidenceHalfWidth 1.93 ### Source [Original source](https://labs.scale.com/leaderboard/fortress) Source scale Original model label gpt-oss-120b Observed 2026-10-21T16:35:37.552Z Evidence SHA-256 `875f1c5b1112c4b84f3d17df72771ebee5c293c542cbb8ab3399e35c4f82ffa5` Source locator `/rows/1` | | | | | --- Cite as: Fortress — Benchize, https://benchize.com/benchmarks/scale%3Afortress (retrieved 2026-10-21)