GPT-6 Sol security benchmark rundown
GPT-6 Sol (OpenAI) ranks 10th of 26 on BlueBench with 71.1 overall at $1.74 per task. Track scores, value against peers, strengths, weaknesses, and when to use it.
Verdict
GPT-6 Sol ranks 10th of 26 on BlueBench with an overall score of 71.1, in the upper half of the leaderboard and 12.1 points behind Claude Opus 5.5. At $1.74 per task it is the 15th cheapest of 26, and Claude Sonnet 5.5 scores higher (77.5) for less ($1.01). Against the 12 models priced within 2x of its cost per task, it scores within 3 points of the peer median overall; it is strongest against them in incident response and weakest in threat hunting. Malware Analysis, weighted 10%, is a secondary weak spot (−7.3 vs the peer median). Its results are among the most consistent across the benchmarks (reliability 46.3, 7th of 26).
- BlueBench score
- 71.1
- #10 of 26 models
- Cost per task
- $1.74
- #15 cheapest of 26
- Reliability
- 46.3
- #7 of 26 models
- Overall vs cost peers
- +2.5
- Off the cost–score frontier
Capability by track
Incident response, threat hunting, and detection engineering each carry 30% of the BlueBench score; malware analysis carries 10%.
GPT-6 Sol scores 71.1 overall, 10th of 26 on the capability leaderboard. Its best rank is in incident response (77.9, 9th of 26) and its worst is in threat hunting (56.2, 16th of 26).
| Track | Weight | Score | Rank | Peer median | vs peers |
|---|---|---|---|---|---|
| Incident Response | 30% | 77.9 | #9 of 26 | 68.8 | +9.2 |
| Threat Hunting | 30% | 56.2 | #16 of 26 | 61.9 | −5.7 |
| Detection Engineering | 30% | 78.4 | #11 of 26 | 76.7 | +1.8 |
| Malware Analysis | 10% | 73.8 | #16 of 26 | 81.1 | −7.3 |
Cost, reliability, and value
Cost is USD per evaluated task including repeated attempts. Reliability is 100 minus the mean score range across the six benchmarks.
GPT-6 Sol averages $1.74 per task, including repeated attempts: the 15th cheapest of 26 models with a cost figure. It is off the cost–score frontier: Claude Sonnet 5.5 scores 77.5 at $1.01, and 3 other models are also both cheaper and at least as good. Its peer group is the 12 models priced within 2x of its cost per task. Their median overall score is 68.6, so GPT-6 Sol is in line with its price range (+2.5). Reliability, 100 minus its mean score range across the benchmarks, is 46.3 (7th of 26).
Cost peers: the 12 models priced within 2x of its cost per task
Strengths and weaknesses
Each track compared with the median of the 12 models priced within 2x of its cost per task. A gap of 5 points or more counts.
Against the 12 models priced within 2x of its cost per task, GPT-6 Sol is strongest in incident response (+9.2). It lags them in threat hunting (−5.7) and malware analysis (−7.3). Differences are versus the peer median score in each track; ±5 points is the threshold.
Beats its cost peers
- Incident Response77.9 (+9.2)
Lags its cost peers
- Threat Hunting56.2 (−5.7)
- Malware Analysis73.8 (−7.3)
Nearest neighbors on the leaderboard
The models ranked just above and below it overall, with the largest track gaps.
- Claude Opus 4.8 sits 2 places above, 3.8 points higher overall (74.9), at about the same cost per task ($1.56); it is 12.4 points ahead in threat hunting and 9.5 points behind in incident response.
- DeepSeek V4.1 Flash sits one place above, 0.4 points higher overall (71.5), at 10% of the cost per task ($0.17); it is 12.3 points ahead in threat hunting and 17.0 points behind in incident response.
- GLM-5.3 sits one place below, 0.2 points lower overall (70.9), at 1.9x the cost per task ($3.27); it is 12.0 points ahead in threat hunting and 14.3 points behind in detection engineering.
- GLM-5.3 Flash sits 2 places below, 0.5 points lower overall (70.6), at 6% of the cost per task ($0.11); it is 7.0 points ahead in threat hunting and 5.3 points behind in incident response.
| Rank | Model | Score | $/task |
|---|---|---|---|
| 8 | Claude Opus 4.8 | 74.9 | $1.56 |
| 9 | DeepSeek V4.1 Flash | 71.5 | $0.17 |
| 10 | GPT-6 Sol | 71.1 | $1.74 |
| 11 | GLM-5.3 | 70.9 | $3.27 |
| 12 | GLM-5.3 Flash | 70.6 | $0.11 |
Best for
Alternatives are limited to models costing at most 1.5x as much, except the overall leader and the cheaper-model pick.
- Use itUse it for incident response (77.9, +9.2) at its price: that track beats the median of the 12 models priced within 2x of its cost per task by 5 points or more.
- InsteadUse Claude Sonnet 5.5 instead in most cases: it scores 77.5 overall (+6.4) at $1.01 per task (58% of the cost per task).
- InsteadUse Claude Opus 4.8 instead if threat hunting is the priority: it scores 68.6 there (vs 56.2) at $1.56 per task.
- InsteadUse Claude Opus 5.5 instead if the highest overall score matters more than cost: it scores 83.2 (12.1 points higher) at $3.54 per task.
Notable benchmark results
Each BlueBench benchmark is a separate intrusion, real or simulated, with its own tasks and grading.
- Its best standing is on Simulation-002 (Identity & Active Directory Attacks): 79.5 (6th of 26).
- Its scores range from 50.0 on Simulation-001 to 92.3 on Intrusion-001, a spread of 42.2 points.
| Benchmark | Score | Rank |
|---|---|---|
| Simulation-003Impact & Exfiltration | 70.0 | #10 of 26 |
| Simulation-002Identity & Active Directory Attacks | 79.5 | #6 of 26 |
| Simulation-001Initial Access & Command-and-Control | 50.0 | #16 of 26 |
| Intrusion-003AWS Cloud Intrusion | 79.4 | #11 of 26 |
| Intrusion-002Windows Enterprise Intrusion | 68.7 | #14 of 26 |
| Intrusion-001macOS Threat Investigation | 92.3 | #8 of 26 |
About this rundown
- Provider
- OpenAI · closed weight
- Tested
- Sep 30, 2026 – Oct 1, 2026
- Suite
- 74 tasks across 6 benchmarks, evaluated in a common harness
- Data updated
- Oct 27, 2026
- How it is written
- Generated from the BlueBench result files with fixed rules: cost peers are models priced within 2x per task (or the nearest by cost when fewer than three are), a track is a strength or weakness at 5+ points from the peer median, and ranks share a place on ties. The text is regenerated whenever the results change.
AI for the blue team.
Run Cotool's harness in your environment to get real security work done