Gemini 4 Argon security benchmark rundown
Gemini 4 Argon (Google) ranks 3rd of 26 on BlueBench with 79.9 overall at $6.76 per task. Track scores, value against peers, strengths, weaknesses, and when to use it.
Verdict
Gemini 4 Argon ranks 3rd of 26 on BlueBench with an overall score of 79.9, in the top quarter of the leaderboard and 3.3 points behind Claude Opus 5.5. At $6.76 per task it is the 25th cheapest of 26, and Claude Opus 5.5 scores higher (83.2) for less ($3.54). Against the 5 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 threat hunting. Malware Analysis, weighted 10%, is a secondary weak spot (−11.3 vs the peer median).
- BlueBench score
- 79.9
- #3 of 26 models
- Cost per task
- $6.76
- #25 cheapest of 26
- Reliability
- 38.0
- #12 of 26 models
- Overall vs cost peers
- +0.7
- 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%.
Gemini 4 Argon scores 79.9 overall, 3rd of 26 on the capability leaderboard. Its best rank is in threat hunting (74.5, 2nd of 26) and its worst is in malware analysis (72.5, 18th of 26).
| Track | Weight | Score | Rank | Peer median | vs peers |
|---|---|---|---|---|---|
| Incident Response | 30% | 86.8 | #3 of 26 | 80.8 | +6.0 |
| Threat Hunting | 30% | 74.5 | #2 of 26 | 62.8 | +11.7 |
| Detection Engineering | 30% | 80.7 | #8 of 26 | 82.6 | −1.9 |
| Malware Analysis | 10% | 72.5 | #18 of 26 | 83.8 | −11.3 |
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.
Gemini 4 Argon averages $6.76 per task, including repeated attempts: the 25th 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 5 models priced within 2x of its cost per task. Their median overall score is 79.2, so Gemini 4 Argon is in line with its price range (+0.7). Reliability, 100 minus its mean score range across the benchmarks, is 38.0 (12th of 26).
Cost peers: the 5 models priced within 2x of its cost per task
Strengths and weaknesses
Each track compared with the median of the 5 models priced within 2x of its cost per task. A gap of 5 points or more counts.
Against the 5 models priced within 2x of its cost per task, Gemini 4 Argon is strongest in threat hunting (+11.7) and incident response (+6.0). It lags them in malware analysis (−11.3). Differences are versus the peer median score in each track; ±5 points is the threshold.
Beats its cost peers
- Threat Hunting74.5 (+11.7)
- Incident Response86.8 (+6.0)
Lags its cost peers
- Malware Analysis72.5 (−11.3)
Nearest neighbors on the leaderboard
The models ranked just above and below it overall, with the largest track gaps.
- Claude Opus 5.5 sits 2 places above, 3.3 points higher overall (83.2), at 52% of the cost per task ($3.54); it is 16.9 points ahead in malware analysis.
- GPT-6 Astra sits one place above, 0.4 points higher overall (80.3), at 1.4x the cost per task ($9.60); it is 16.3 points ahead in malware analysis and 11.7 points behind in threat hunting.
- Grok 4.6 sits one place below, 0.7 points lower overall (79.2), at 54% of the cost per task ($3.64); it is 11.3 points ahead in malware analysis and 6.0 points behind in incident response.
- Claude Sonnet 5.5 sits 2 places below, 2.4 points lower overall (77.5), at 15% of the cost per task ($1.01); it is 15.3 points ahead in malware analysis and 7.6 points behind in threat hunting.
| Rank | Model | Score | $/task |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 83.2 | $3.54 |
| 2 | GPT-6 Astra | 80.3 | $9.60 |
| 3 | Gemini 4 Argon | 79.9 | $6.76 |
| 4 | Grok 4.6 | 79.2 | $3.64 |
| 5 | Claude Sonnet 5.5 | 77.5 | $1.01 |
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 (74.5, 2nd of 26) and incident response (86.8, 3rd of 26): each is a top-3 result among all models.
- InsteadUse Claude Opus 5.5 instead in most cases: it scores 83.2 overall (+3.3) at $3.54 per task (52% of the cost per task).
Notable benchmark results
Each BlueBench benchmark is a separate intrusion, real or simulated, with its own tasks and grading.
- 2nd place on Simulation-002 (Identity & Active Directory Attacks): 86.6 (2nd of 26).
- 2nd place on Simulation-001 (Initial Access & Command-and-Control): 72.8 (2nd of 26).
- Its scores range from 72.8 on Simulation-001 to 91.9 on Intrusion-001, a spread of 19.1 points.
- 2 recorded refusals across the benchmarks; refused attempts stay in the score.
| Benchmark | Score | Rank |
|---|---|---|
| Simulation-003Impact & Exfiltration | 79.3 | #7 of 26 |
| Simulation-002Identity & Active Directory Attacks | 86.6 | #2 of 26 |
| Simulation-001Initial Access & Command-and-Control | 72.8 | #2 of 26 |
| Intrusion-003AWS Cloud Intrusion | 84.6 | #6 of 26 |
| Intrusion-002Windows Enterprise Intrusion | 72.9 | #8 of 26 |
| Intrusion-001macOS Threat Investigation | 91.9 | #9 of 26 |
About this rundown
- Provider
- Google · closed weight
- Tested
- Oct 14, 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