Skip to main content

GPT-6.1 Terra security benchmark rundown

GPT-6.1 Terra (OpenAI) ranks 13th of 26 on BlueBench with 70.5 overall at $0.16 per task. Track scores, value against peers, strengths, weaknesses, and when to use it.

OpenAI · Closed weight · Tested Oct 27, 2026 · Data updated Oct 2026
Methodology

Verdict

GPT-6.1 Terra ranks 13th of 26 on BlueBench with an overall score of 70.5, in the upper half of the leaderboard and 12.7 points behind Claude Opus 5.5. At $0.16 per task it is the 2nd cheapest of 26, and GLM-5.3 Flash scores higher (70.6) for less ($0.11). Against the 4 models priced within 2x of its cost per task, it scores 3.2 points above the peer median overall; it is strongest against them in incident response.

BlueBench score
70.5
#13 of 26 models
Cost per task
$0.16
#2 cheapest of 26
Reliability
45.1
#9 of 26 models
Overall vs cost peers
+3.2
Off the cost–score frontier
All GPT-6.1 Terra results and FAQ →

Capability by track

Incident response, threat hunting, and detection engineering each carry 30% of the BlueBench score; malware analysis carries 10%.

GPT-6.1 Terra scores 70.5 overall, 13th of 26 on the capability leaderboard. Its best rank is in incident response (78.2, 8th of 26) and its worst is in detection engineering (72.5, 17th of 26).

TrackWeightScoreRankPeer medianvs peers
Incident Response30%78.2#8 of 2666.0+12.2
Threat Hunting30%59.6#15 of 2657.5+2.1
Detection Engineering30%72.5#17 of 2672.3+0.2
Malware Analysis10%74.1#15 of 2670.0+4.1

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-6.1 Terra averages $0.16 per task, including repeated attempts: the 2nd cheapest of 26 models with a cost figure. It is off the cost–score frontier: GLM-5.3 Flash scores 70.6 at $0.11. Its peer group is the 4 models priced within 2x of its cost per task. Their median overall score is 67.3, so GPT-6.1 Terra is 3.2 points ahead of its price range. Reliability, 100 minus its mean score range across the benchmarks, is 45.1 (9th of 26).

Strengths and weaknesses

Each track compared with the median of the 4 models priced within 2x of its cost per task. A gap of 5 points or more counts.

Against the 4 models priced within 2x of its cost per task, GPT-6.1 Terra is strongest in incident response (+12.2). Differences are versus the peer median score in each track; ±5 points is the threshold.

Beats its cost peers

  • Incident Response78.2 (+12.2)

Lags its cost peers

No track by 5+ points

Nearest neighbors on the leaderboard

The models ranked just above and below it overall, with the largest track gaps.

  • GLM-5.3 sits 2 places above, 0.4 points higher overall (70.9), at 19.8x the cost per task ($3.27); it is 8.6 points ahead in threat hunting and 8.4 points behind in detection engineering.
  • GLM-5.3 Flash sits one place above, 0.1 points higher overall (70.6), at 65% of the cost per task ($0.11); it is 5.6 points behind in incident response.
  • GPT-5.6 Terra sits one place below, 0.6 points lower overall (69.9), at 6.1x the cost per task ($1.01); it is 9.1 points behind in incident response.
  • Claude Haiku 5.5 sits 2 places below, 2.0 points lower overall (68.5), at 3.6x the cost per task ($0.59); it is 8.0 points ahead in detection engineering and 12.7 points behind in incident response.
RankModelScore$/task
11GLM-5.370.9$3.27
12GLM-5.3 Flash70.6$0.11
13GPT-6.1 Terra70.5$0.16
14GPT-5.6 Terra69.9$1.01
15Claude Haiku 5.568.5$0.59

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 (78.2, +12.2) at its price: that track beats the median of the 4 models priced within 2x of its cost per task by 5 points or more.
  • InsteadUse GLM-5.3 Flash instead in most cases: it scores 70.6 overall (+0.1) at $0.11 per task (65% of the cost per task).
  • InsteadUse DeepSeek V4.1 Flash instead if detection engineering is the priority: it scores 81.7 there (vs 72.5) at $0.17 per task.
  • InsteadUse Claude Opus 5.5 instead if the highest overall score matters more than cost: it scores 83.2 (12.7 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 Intrusion-003 (AWS Cloud Intrusion): 84.0 (7th of 26).
  • Its scores range from 49.0 on Simulation-001 to 85.4 on Intrusion-001, a spread of 36.4 points.
BenchmarkScoreRank
Simulation-003Impact & Exfiltration74.3#8 of 26
Simulation-002Identity & Active Directory Attacks73.3#9 of 26
Simulation-001Initial Access & Command-and-Control49.0#17 of 26
Intrusion-003AWS Cloud Intrusion84.0#7 of 26
Intrusion-002Windows Enterprise Intrusion66.9#17 of 26
Intrusion-001macOS Threat Investigation85.4#20 of 26

About this rundown

Provider
OpenAI · closed weight
Tested
Oct 27, 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

Book a demo