Skip to main content

Claude Sonnet 5 security benchmark rundown

Claude Sonnet 5 (Anthropic) ranks 26th of 26 on BlueBench with 44.8 overall at $1.77 per task. Track scores, value against peers, strengths, weaknesses, and when to use it.

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

Verdict

Claude Sonnet 5 ranks 26th of 26 on BlueBench with an overall score of 44.8, in the bottom quarter of the leaderboard and 38.4 points behind Claude Opus 5.5. At $1.77 per task it is the 16th 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 25.6 points below the peer median overall; it trails them most in incident response. Malware Analysis, weighted 10%, is a secondary strength (+8.2 vs the peer median). Its results vary more than most across the benchmarks (reliability 8.1, 25th of 26). It refused a material share of tasks on cybersecurity grounds, and those refusals count against its scores.

BlueBench score
44.8
#26 of 26 models
Cost per task
$1.77
#16 cheapest of 26
Reliability
8.1
#25 of 26 models
Overall vs cost peers
−25.6
Off the cost–score frontier
All Claude Sonnet 5 results and FAQ →

Capability by track

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

Claude Sonnet 5 scores 44.8 overall, 26th of 26 on the capability leaderboard. Its best rank is in malware analysis (86.9, 6th of 26) and its worst is in threat hunting (23.8, 26th of 26).

TrackWeightScoreRankPeer medianvs peers
Incident Response30%29.2#26 of 2670.7−41.5
Threat Hunting30%23.8#26 of 2661.9−38.1
Detection Engineering30%67.3#20 of 2677.8−10.5
Malware Analysis10%86.9#6 of 2678.8+8.2

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.

Claude Sonnet 5 averages $1.77 per task, including repeated attempts: the 16th 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 14 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 70.4, so Claude Sonnet 5 is 25.6 points behind its price range. Reliability, 100 minus its mean score range across the benchmarks, is 8.1 (25th of 26).

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, Claude Sonnet 5 is strongest in malware analysis (+8.2). It lags them in incident response (−41.5), threat hunting (−38.1) and detection engineering (−10.5). Differences are versus the peer median score in each track; ±5 points is the threshold.

Beats its cost peers

  • Malware Analysis86.9 (+8.2)

Lags its cost peers

  • Incident Response29.2 (−41.5)
  • Threat Hunting23.8 (−38.1)
  • Detection Engineering67.3 (−10.5)

Nearest neighbors on the leaderboard

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

  • Claude Opus 5 sits 2 places above, 5.7 points higher overall (50.5), at 2.5x the cost per task ($4.44); it is 12.3 points ahead in threat hunting and 32.2 points behind in malware analysis.
  • Gemini 3.1 Pro sits one place above, 0.5 points higher overall (45.3), at 2.6x the cost per task ($4.63); it is 19.4 points ahead in threat hunting and 28.6 points behind in malware analysis.
RankModelScore$/task
24Claude Opus 550.5$4.44
25Gemini 3.1 Pro45.3$4.63
26Claude Sonnet 544.8$1.77

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 malware analysis (86.9, +8.2) 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 (+32.7) at $1.01 per task (57% of the cost per task).
  • InsteadUse GPT-6.1 Sol instead if incident response is the priority: it scores 86.4 there (vs 29.2) at $1.30 per task.
  • InsteadUse Claude Opus 5.5 instead if the highest overall score matters more than cost: it scores 83.2 (38.4 points higher) at $3.54 per task.
  • CautionTest it on your own workflows before relying on it: it refused a material share of BlueBench tasks on cybersecurity grounds.

Notable benchmark results

Each BlueBench benchmark is a separate intrusion, real or simulated, with its own tasks and grading.

  • Near the bottom on Simulation-003 (Impact & Exfiltration): 47.8 (24th of 26).
  • Near the bottom on Simulation-001 (Initial Access & Command-and-Control): 36.4 (24th of 26).
  • Near the bottom on Intrusion-003 (AWS Cloud Intrusion): 28.0 (26th of 26).
  • Near the bottom on Intrusion-002 (Windows Enterprise Intrusion): 51.0 (24th of 26).
  • Near the bottom on Intrusion-001 (macOS Threat Investigation): 45.8 (26th of 26).
  • Its scores range from 28.0 on Intrusion-003 to 51.0 on Intrusion-002, a spread of 22.9 points.
  • 3 recorded refusals across the benchmarks; refused attempts stay in the score.
BenchmarkScoreRank
Simulation-003Impact & Exfiltration47.8#24 of 26
Simulation-002Identity & Active Directory Attacks44.8#22 of 26
Simulation-001Initial Access & Command-and-Control36.4#24 of 26
Intrusion-003AWS Cloud Intrusion28.0#26 of 26
Intrusion-002Windows Enterprise Intrusion51.0#24 of 26
Intrusion-001macOS Threat Investigation45.8#26 of 26

Notes

Written by hand; they only restate the data above.

  • Claude Sonnet 5 scores 67.3 in detection engineering and 86.9 in malware analysis, but 29.2 in incident response and 23.8 in threat hunting, the two tracks graded against a hidden rubric from a written report.

About this rundown

Provider
Anthropic · 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