Skip to main content

Grok 4.6 security benchmark rundown

Grok 4.6 (xAI) ranks 4th of 26 on BlueBench with 79.2 overall at $3.64 per task. Track scores, value against peers, strengths, weaknesses, and when to use it.

xAI · Closed weight · Tested Sep 11, 2026 · Data updated Oct 2026
Methodology

Verdict

Grok 4.6 ranks 4th of 26 on BlueBench with an overall score of 79.2, in the top quarter of the leaderboard and 4.0 points behind Claude Opus 5.5. At $3.64 per task it is the 22nd cheapest of 26, and Claude Opus 5.5 scores higher (83.2) for less ($3.54). Against the 8 models priced within 2x of its cost per task, it scores 10.1 points above the peer median overall; it is strongest against them in threat hunting. Malware Analysis, weighted 10%, is a secondary strength (+9.0 vs the peer median). Its results are among the most consistent across the benchmarks (reliability 47.4, 6th of 26).

BlueBench score
79.2
#4 of 26 models
Cost per task
$3.64
#22 cheapest of 26
Reliability
47.4
#6 of 26 models
Overall vs cost peers
+10.1
Off the cost–score frontier
All Grok 4.6 results and FAQ →

Capability by track

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

Grok 4.6 scores 79.2 overall, 4th of 26 on the capability leaderboard. Its best rank is in threat hunting (72.6, 3rd of 26) and its worst is in incident response (80.8, 7th of 26).

TrackWeightScoreRankPeer medianvs peers
Incident Response30%80.8#7 of 2674.8+6.0
Threat Hunting30%72.6#3 of 2663.5+9.1
Detection Engineering30%82.6#5 of 2678.5+4.1
Malware Analysis10%83.8#7 of 2674.8+9.0

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.

Grok 4.6 averages $3.64 per task, including repeated attempts: the 22nd cheapest of 26 models with a cost figure. It is off the cost–score frontier: Claude Opus 5.5 scores 83.2 at $3.54. Its peer group is the 8 models priced within 2x of its cost per task. Their median overall score is 69.1, so Grok 4.6 is 10.1 points ahead of its price range. Reliability, 100 minus its mean score range across the benchmarks, is 47.4 (6th of 26).

Strengths and weaknesses

Each track compared with the median of the 8 models priced within 2x of its cost per task. A gap of 5 points or more counts.

Against the 8 models priced within 2x of its cost per task, Grok 4.6 is strongest in threat hunting (+9.1), incident response (+6.0) and malware analysis (+9.0). Differences are versus the peer median score in each track; ±5 points is the threshold.

Beats its cost peers

  • Threat Hunting72.6 (+9.1)
  • Incident Response80.8 (+6.0)
  • Malware Analysis83.8 (+9.0)

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.

  • GPT-6 Astra sits 2 places above, 1.1 points higher overall (80.3), at 2.6x the cost per task ($9.60); it is 7.2 points ahead in incident response and 9.8 points behind in threat hunting.
  • Gemini 4 Argon sits one place above, 0.7 points higher overall (79.9), at 1.9x the cost per task ($6.76); it is 6.0 points ahead in incident response and 11.3 points behind in malware analysis.
  • Claude Sonnet 5.5 sits one place below, 1.7 points lower overall (77.5), at 28% of the cost per task ($1.01); it is 5.7 points behind in threat hunting.
  • GPT-5.6 Sol sits 2 places below, 2.0 points lower overall (77.2), at 79% of the cost per task ($2.89); it is 9.9 points behind in threat hunting.
RankModelScore$/task
2GPT-6 Astra80.3$9.60
3Gemini 4 Argon79.9$6.76
4Grok 4.679.2$3.64
5Claude Sonnet 5.577.5$1.01
6GPT-5.6 Sol77.2$2.89

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 threat hunting (72.6, 3rd of 26): it is a top-3 result among all models.
  • Use itUse it for incident response (80.8, +6.0) and malware analysis (83.8, +9.0) at its price: those tracks beat the median of the 8 models priced within 2x of its cost per task by 5 points or more.
  • InsteadUse Claude Opus 5.5 instead in most cases: it scores 83.2 overall (+4.0) at $3.54 per task (about the same cost 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-001 (Initial Access & Command-and-Control): 64.9 (4th of 26).
  • Its scores range from 64.9 on Simulation-001 to 97.5 on Intrusion-001, a spread of 32.5 points.
BenchmarkScoreRank
Simulation-003Impact & Exfiltration80.6#6 of 26
Simulation-002Identity & Active Directory Attacks80.8#5 of 26
Simulation-001Initial Access & Command-and-Control64.9#4 of 26
Intrusion-003AWS Cloud Intrusion87.4#4 of 26
Intrusion-002Windows Enterprise Intrusion72.1#10 of 26
Intrusion-001macOS Threat Investigation97.5#4 of 26

About this rundown

Provider
xAI · closed weight
Tested
Sep 11, 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