IT AI Index
September 2026 Edition · The permanent record of this edition. The unqualified address always carries the latest edition.
Index Security operations September 2026 Edition

Penetration testing as a service

Asked as “penetration testing service”, and as “pentest-as-a-service provider”, on behalf of a mid-market B2B company. 45 first choices recorded across the direct, paraphrase, budget and scale prompts, twelve models each.
Standing
Clear leader
44% of first choices, clear leader.

By buyer segment

The same question asked on behalf of a different buyer. Each standing is computed within its segment; they sit side by side and are never added together.

01The standing

Share is the count of first choices across the direct, paraphrase, budget and scale prompts, over all twelve models, for a mid-market B2B company. Ordered by share.
ProductFirst-choice shareNegative rateLabelsQuadrant
01Cobalt44%5%40endorsed leader
02Astra Security13%0%26accepted challenger
03BreachLock9%3%30accepted challenger
04NetSPI2%9%22accepted challenger
Show the four products at 0%, ordered by negative rate
07Bugcrowd0%30%10criticized challenger
08HackerOne0%30%10criticized challenger
06Synack0%11%18accepted challenger
05Bishop Fox0%6%16accepted challenger
Bars are the share of first choices, 0 to 100Every product with at least 10 labels here. Every product name links to its vendor page.

Recommended versus criticized

Every product with at least 10 labels here, on both axes. The 30% line names a quadrant, not the verdict above: that one needs more than 40%.

Criticized challengerCriticized default
Negative label rate →
01
02
03
04
05
06
07
08
Accepted challengerEndorsed leader
0%First-choice share → · lines at 30% share and 25% negative60%
Key
01Cobalt44%
02Astra Security13%
03BreachLock9%
04NetSPI2%
05Bishop Fox0%
06Synack0%
07Bugcrowd0%
08HackerOne0%

02What they warned about

Five of twelve models held their first choice under the paraphrase. Claude Haiku 4.5, Grok 4.1 Fast, Mistral Small, Llama 4 Maverick, Qwen 3.7 Flash, GLM 4.7 FlashX and MiniMax M2.5 changed. A high negative share on a product with few labels is a warning. A low share on a product with many labels is salience, not sentiment.
HackerOne
30%
3 of 10 labels negative · 3 of 12 models
“While good for finding unique bugs, tester continuity is lower because you never know which specific hacker will test your system next.” Qwen 3.7 Flash, comparative prompt
Acunetix
33%
2 of 6 labels negative · 2 of 12 models
“but it is more of a web app security product than a full manual pentest firm” Perplexity Sonar, budget prompt
Bugcrowd
30%
3 of 10 labels negative · 2 of 12 models
“typically more expensive for enterprise‑scale needs; often better suited to large, complex programs” GLM 4.7 FlashX, paraphrase prompt
Cobalt
5%
2 of 40 labels negative · 2 of 12 models
“though it is more of a managed marketplace than a low-cost commodity service” Perplexity Sonar, budget prompt

03What they cite

Citations exist only for the models that return a source list: twelve of the twelve in this edition, and all six flagship models on the expanded tier.

Sites the answers cite

62 of 72 answers in this category came back with a source list, from 12 of 12 models: citations where the model returns them, or the search results it consulted. 6 of those lists are Google grounding redirects that name no site and are left out of the counts. 740 links across 184 sites, every framing counted. Ranked by the number of answers carrying the site or page.

vendor site · DeepStrike37 answers · 64 citations · 10 models
vendor site · Synack30 answers · 39 citations · 11 models
vendor site · UnderDefense21 answers · 24 citations · 9 models
20 answers · 21 citations · 9 models
vendor site · Stingrai17 answers · 24 citations · 9 models
15 answers · 22 citations · 9 models
15 answers · 21 citations · 9 models
vendor site · Software Secured15 answers · 17 citations · 8 models
vendor site · Cybri13 answers · 15 citations · 9 models
13 answers · 13 citations · 9 models
12 answers · 12 citations · 10 models
11 answers · 15 citations · 7 models

Pages the answers cite

The ten pages named in the most answers, by full address. A page here is one the models returned with a recommendation, not one the index endorses.

04How they answered

Six framings of the same buying question, each sent to every model in a fresh session with search on. One row per model, so a row shows whether it held its answer under rewording, what it named when cost was the constraint, and what it argued against. Computed from the raw judge labels.
ModelDirect“What is the best penetration testing service for a mid-market B2B company?”Paraphrase“Which pentest-as-a-service provider would you recommend to a mid-sized B2B company?”Comparative“What are the top penetration testing services and how do they differ?”Budget-constrained“What is the best penetration testing service for a company with a limited budget?”Scale-constrained“We are a 500 person company evaluating a penetration testing service. What should we look at?”Negative“Which penetration testing services should I avoid or be cautious about?”
Claude Haiku 4.5no first choiceBreachLockChanged
Four alternativesCobalt, NetSPI, Packetlabs, Software Secured
Rapid7
Eight alternativesBeagle Security, Bishop Fox, Coalfire, Cobalt, IBM X-Force Red, NetSPI, Synack, Trustwave
BreachLock, Kualitatem, UnderDefense
One alternativeCyber Protection Group
against: Fiverr
no first choicenothing named
GPT-5.4 miniCobalt
Six alternativesAstra Security, HackerOne, Packetlabs, Thoropass Pentesting, UnderDefense, Vumetric
CobaltHeld
Two alternativesBugcrowd, Synack
no first choiceno first choiceno first choicenothing named
Gemini 3.5 FlashCobalt
Three alternativesBreachLock, Cybri, NetSPI
CobaltHeld
Four alternativesAstra Security, BreachLock, NetSPI, Stingrai
no first choice
Seven alternativesBishop Fox, BreachLock, Cobalt, Horizon3.ai NodeZero, NetSPI, Pentera, Synack
Vonahi Security vPenTest
Four alternativesAstra Security, Beagle Security, Cyber Protection Group, TCM Security
no first choicenothing named
Perplexity SonarCobalt
Five alternativesBreachLock, DeepStrike, Prescient Security, Stingrai, UnderDefense
CobaltHeld
Three alternativesBreachLock, DeepStrike, Sprocket Security
no first choiceAstra Security
Three alternativesBreachLock, Qualysec, UnderDefense
against: Acunetix, Cobalt, Intruder
no first choicenothing named
Grok 4.1 FastPacketlabs
Two alternativesBreachLock, Cobalt
against: NetSPI, Synack
CobaltChanged
One alternativeBreachLock
NetSPI
Five alternativesAstra Security, Bishop Fox, BreachLock, Cobalt, Synack
against: Rapid7
Bug Circuit
Three alternativesAffordable Pentesting, Astra Security, Practical Assurance
against: BreachLock, Cobalt
no first choicenothing named
Mistral SmallCobalt
Three alternativesBreachLock, Packetlabs, Stingrai
Astra Security, CobaltChanged
Two alternativesBreachLock, Software Secured
Astra Security
Ten alternativesBishop Fox, BreachLock, Cobalt, DeepStrike, Intruder, Invicti, NetSPI, Rapid7, Synack, White Knight Labs
Astra Security, Cobalt
Two alternativesAcunetix, Intruder
no first choicenothing named
DeepSeek V4 FlashCobalt
Four alternativesAstra Security, BreachLock, NetSPI, Packetlabs
CobaltHeld
Two alternativesAstra Security, BreachLock
Astra Security
Seven alternativesBishop Fox, BreachLock, Bugcrowd, Cobalt, HackerOne, NetSPI, Praetorian
against: Acunetix, Intruder, Invicti
Astra Security, BreachLock
Three alternativesBeagle Security, Intruder, Vonahi Security vPenTest
against: Bug Circuit, TurboPentest
no first choicenothing named
Llama 4 MaverickCobalt
Two alternativesDeepStrike, Echelon Cyber
Software SecuredChangedno first choiceBreachLock, Cobalt
One alternativeAstra Security
no first choiceagainst: Black Hills Information Security, Trustwave
Qwen 3.7 FlashCobalt
Four alternativesCoalfire, Kforce, Synack, Trustwave
against: Bugcrowd, CrowdStrike, Mandiant
Astra SecurityChanged
Two alternativesCobalt, Software Secured
Coalfire, Cobalt, Intruder, NetSPI
Two alternativesBreachLock, SpecterOps
against: Bugcrowd, HackerOne
BreachDefense, Luta Security
Three alternativesBugcrowd, HackerOne, Intigriti
against: Deloitte, EY, KPMG, OpenVAS, PwC, Tenable Nessus
no first choiceagainst: Deloitte, KPMGnothing named
Kimi K2Cobalt
Three alternativesBreachLock, NetSPI, UnderDefense
against: Bishop Fox
CobaltHeld
Two alternativesAstra Security, Synack
against: HackerOne
Astra Security
Four alternativesBishop Fox, BreachLock, Cobalt, Optiv
Bug Circuit, HIFENCE
Three alternativesAcunetix, BreachLock, Cobalt
no first choicenothing named
GLM 4.7 FlashXCobalt, NetSPI
Five alternativesAstra Security, Bishop Fox, BreachLock, Secureworks, Synack
CobaltChanged
Two alternativesBreachLock, Sprocket Security
against: Bugcrowd, HackerOne, NetSPI, Rapid7, Synack
Bishop Fox, Bugcrowd, NetSPI
Two alternativesDeepStrike, Synack
Bug Circuit, TurboPentest
One alternativeAffordable Pentesting
no first choicenothing named
MiniMax M2.5Stingrai
Two alternativesCobalt, Invadel
Astra Security, CobaltChanged
Two alternativesBreachLock, Terra Security
Astra Security
Four alternativesAcunetix, Cobalt, Intruder, Invicti
Affordable Pentesting
Three alternativesBreachLock, Cobalt, CyberSuite
no first choice
Five alternativesAstra Security, Bishop Fox, Cobalt, NetSPI, Synack
nothing named
Bold is the first choiceAlternatives are counted; the count opens them.What the answer argued against

05The record

One row per call: the version string exactly as returned, whether the model searched, sources cited, and latency. Full answer text is in the free responses file. Download the record
Seventy-two rows: every prompt, every model, every answer.
PromptModelVersion stringTime (UTC)SearchedSourcesLatency
Direct recommendationClaude Haiku 4.5claude-haiku-4-5-202510012026-09-17 14:13no04 s
Direct recommendationGPT-5.4 minigpt-5.4-mini-2026-03-172026-09-17 10:58yes36 s
Direct recommendationGemini 3.5 Flashgemini-3.5-flash2026-09-17 10:29yes918 s
Direct recommendationPerplexity Sonarsonar2026-09-17 09:48yes207 s
Direct recommendationGrok 4.1 Fastspacexai/grok-4.1-fast-non-reasoning via vertex2026-09-17 14:04yes3423 s
Direct recommendationMistral Smallmistral/mistral-small via mistral2026-09-17 12:12yes55 s
Direct recommendationDeepSeek V4 Flashdeepseek/deepseek-v4-flash via fireworks2026-09-17 10:54yes914 s
Direct recommendationLlama 4 Maverickmeta/llama-4-maverick via bedrock2026-09-17 09:46yes53 s
Direct recommendationQwen 3.7 Flashalibaba/qwen3.7-flash via alibaba2026-09-17 14:03no044 s
Direct recommendationKimi K2moonshotai/kimi-k2 via novita2026-09-17 13:49yes1247 s
Direct recommendationGLM 4.7 FlashXzai/glm-4.7-flashx via zai2026-09-17 10:43yes1542 s
Direct recommendationMiniMax M2.5minimax/minimax-m2.5 via minimax2026-09-17 10:27yes530 s
ParaphraseClaude Haiku 4.5claude-haiku-4-5-202510012026-09-17 12:25yes179 s
ParaphraseGPT-5.4 minigpt-5.4-mini-2026-03-172026-09-17 09:54yes45 s
ParaphraseGemini 3.5 Flashgemini-3.5-flash2026-09-17 12:56yes2127 s
ParaphrasePerplexity Sonarsonar2026-09-17 13:50yes205 s
ParaphraseGrok 4.1 Fastspacexai/grok-4.1-fast-non-reasoning via vertex2026-09-17 12:54yes3116 s
ParaphraseMistral Smallmistral/mistral-small via mistral2026-09-17 13:03yes55 s
ParaphraseDeepSeek V4 Flashdeepseek/deepseek-v4-flash via fireworks2026-09-17 10:06yes1820 s
ParaphraseLlama 4 Maverickmeta/llama-4-maverick via bedrock2026-09-17 10:50yes52 s
ParaphraseQwen 3.7 Flashalibaba/qwen3.7-flash via alibaba2026-09-17 09:31yes540 s
ParaphraseKimi K2moonshotai/kimi-k2 via novita2026-09-17 11:21yes1944 s
ParaphraseGLM 4.7 FlashXzai/glm-4.7-flashx via zai2026-09-17 10:16yes23111 s
ParaphraseMiniMax M2.5minimax/minimax-m2.5 via minimax2026-09-17 13:41yes539 s
ComparativeClaude Haiku 4.5claude-haiku-4-5-202510012026-09-17 12:20yes1411 s
ComparativeGPT-5.4 minigpt-5.4-mini-2026-03-172026-09-17 10:06yes47 s
ComparativeGemini 3.5 Flashgemini-3.5-flash2026-09-17 11:24yes1825 s
ComparativePerplexity Sonarsonar2026-09-17 12:00yes2010 s
ComparativeGrok 4.1 Fastspacexai/grok-4.1-fast-non-reasoning via vertex2026-09-17 12:17yes2817 s
ComparativeMistral Smallmistral/mistral-small via mistral2026-09-17 11:33yes57 s
ComparativeDeepSeek V4 Flashdeepseek/deepseek-v4-flash via fireworks2026-09-17 13:01yes1724 s
ComparativeLlama 4 Maverickmeta/llama-4-maverick via bedrock2026-09-17 13:48yes53 s
ComparativeQwen 3.7 Flashalibaba/qwen3.7-flash via alibaba2026-09-17 09:59yes1554 s
ComparativeKimi K2moonshotai/kimi-k2 via novita2026-09-17 09:39yes1536 s
ComparativeGLM 4.7 FlashXzai/glm-4.7-flashx via zai2026-09-17 10:07yes24148 s
ComparativeMiniMax M2.5minimax/minimax-m2.5 via minimax2026-09-17 11:00yes535 s
Budget constrainedClaude Haiku 4.5claude-haiku-4-5-202510012026-09-17 14:14yes97 s
Budget constrainedGPT-5.4 minigpt-5.4-mini-2026-03-172026-09-17 11:15no03 s
Budget constrainedGemini 3.5 Flashgemini-3.5-flash2026-09-17 13:05yes2225 s
Budget constrainedPerplexity Sonarsonar2026-09-17 12:30yes206 s
Budget constrainedGrok 4.1 Fastspacexai/grok-4.1-fast-non-reasoning via vertex2026-09-17 10:30yes4020 s
Budget constrainedMistral Smallmistral/mistral-small via mistral2026-09-17 11:54yes96 s
Budget constrainedDeepSeek V4 Flashdeepseek/deepseek-v4-flash via fireworks2026-09-17 14:05yes2824 s
Budget constrainedLlama 4 Maverickmeta/llama-4-maverick via bedrock2026-09-17 12:35yes53 s
Budget constrainedQwen 3.7 Flashalibaba/qwen3.7-flash via alibaba2026-09-17 09:42no036 s
Budget constrainedKimi K2moonshotai/kimi-k2 via novita2026-09-17 10:27yes1020 s
Budget constrainedGLM 4.7 FlashXzai/glm-4.7-flashx via zai2026-09-17 11:24yes26131 s
Budget constrainedMiniMax M2.5minimax/minimax-m2.5 via minimax2026-09-17 09:54yes521 s
Scale constrainedClaude Haiku 4.5claude-haiku-4-5-202510012026-09-17 12:45no05 s
Scale constrainedGPT-5.4 minigpt-5.4-mini-2026-03-172026-09-17 12:39no08 s
Scale constrainedGemini 3.5 Flashgemini-3.5-flash2026-09-17 09:25yes1425 s
Scale constrainedPerplexity Sonarsonar2026-09-17 09:33yes208 s
Scale constrainedGrok 4.1 Fastspacexai/grok-4.1-fast-non-reasoning via vertex2026-09-17 10:47yes1412 s
Scale constrainedMistral Smallmistral/mistral-small via mistral2026-09-17 09:23no08 s
Scale constrainedDeepSeek V4 Flashdeepseek/deepseek-v4-flash via fireworks2026-09-17 10:30yes1017 s
Scale constrainedLlama 4 Maverickmeta/llama-4-maverick via bedrock2026-09-17 13:12yes53 s
Scale constrainedQwen 3.7 Flashalibaba/qwen3.7-flash via alibaba2026-09-17 13:18no037 s
Scale constrainedKimi K2moonshotai/kimi-k2 via novita2026-09-17 09:38yes847 s
Scale constrainedGLM 4.7 FlashXzai/glm-4.7-flashx via zai2026-09-17 11:50yes1483 s
Scale constrainedMiniMax M2.5minimax/minimax-m2.5 via minimax2026-09-17 10:57yes530 s
Negative framingClaude Haiku 4.5claude-haiku-4-5-202510012026-09-17 12:22yes1711 s
Negative framingGPT-5.4 minigpt-5.4-mini-2026-03-172026-09-17 10:02no03 s
Negative framingGemini 3.5 Flashgemini-3.5-flash2026-09-17 12:50yes1518 s
Negative framingPerplexity Sonarsonar2026-09-17 11:36yes208 s
Negative framingGrok 4.1 Fastspacexai/grok-4.1-fast-non-reasoning via vertex2026-09-17 14:03yes1310 s
Negative framingMistral Smallmistral/mistral-small via mistral2026-09-17 12:04yes56 s
Negative framingDeepSeek V4 Flashdeepseek/deepseek-v4-flash via fireworks2026-09-17 11:26yes812 s
Negative framingLlama 4 Maverickmeta/llama-4-maverick via bedrock2026-09-17 12:37yes53 s
Negative framingQwen 3.7 Flashalibaba/qwen3.7-flash via alibaba2026-09-17 13:17yes527 s
Negative framingKimi K2moonshotai/kimi-k2 via novita2026-09-17 11:26yes1530 s
Negative framingGLM 4.7 FlashXzai/glm-4.7-flashx via zai2026-09-17 12:32yes1296 s
Negative framingMiniMax M2.5minimax/minimax-m2.5 via minimax2026-09-17 10:09no010 s

Normalization in this category

Every judgment call made between the raw labels and the numbers above, listed so it is visible and reversible.

Category-scoped readings
Astra read as Astra Security
Vonahi Security (vPenTest) read as Vonahi Security vPenTest
Unresolved, counted raw
Artifice Security
BreachDefense
Breacon
Echelon Cyber
Echelon Risk + Cyber
Fiverr
Kali Linux
Kforce
Luta Security
Penti
Spectral Ops
TCM Security
Thoropass Pentesting
Vumetric
White Knight Labs
Discontinued, still offered
No shut-down product was recommended here.
← MDRAwareness training →