Can AI analyze malware?

It can explain decompiled code, summarize behavioural reports, extract indicators and draft detection rules from material you give it. It does not replace a sandbox or a disassembler, and an analyst has to verify its reading.

Will the model help write malware?

That is not what the service is for. The acceptable use policy does not allow building or spreading malicious code for use on systems you are not authorized to test. The use case here is analysis and detection.

Can it write YARA rules?

Yes. Give it the strings and behaviours you identified and it drafts a rule with its reasoning. Test the rule against clean and malicious sets before deploying it, because a draft can be too broad or too narrow.

How large a file can it read?

It reads text, so the limit is the context window of the build. The largest listed today is 1M tokens; each model page shows its own window.

Per token, from a prepaid balance with a $10 minimum top-up. On 2026-11-11 input prices for security builds start at $0.40 per million tokens.

AI for malware analysis

Malware analysis AI is a language model that helps an analyst understand a sample: it explains decompiled functions, summarizes sandbox reports, extracts indicators and drafts detection rules. PentestLLM provides 9 security-tuned builds by API for that analysis and detection work; it is not a service for building or deploying malware.

Get a key Prices per token

shell

curl https://api.pentestllm.ai/v1/messages \
  -H "x-api-key: sk-ptl-live-…" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "auto/security",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Draft a detection rule for this log pattern and explain each condition."}]
  }'

Rows of server cabinets in a data centre

What the models do for an analyst

Static triage: Read strings, imports and section data you extracted and say what kind of sample this is likely to be and what to look at next.
Code explanation: Explain a decompiled or disassembled function in plain language and suggest names for it and its variables.
Sandbox reports: Condense a long behavioural report into what the sample did, in order, with the lines that support each claim.
Indicators: Pull hashes, domains, paths, mutexes and registry keys out of your notes into a clean list.
Detection rules: Draft YARA and Sigma rules from the behaviour and strings you identified, with a note on false positive risk.
ATT&CK mapping: Map observed behaviour to MITRE ATT&CK techniques for the report.

Why analysts need a model that engages

Malware analysis is defensive work that requires staring at offensive code. A general assistant frequently declines the moment it sees a decompiled routine that touches credentials or encryption, even though the analyst’s goal is to detect and remove it. That makes stock models unreliable in the one place an analyst needs patience.

The security builds here are tuned or abliterated to answer security questions other models decline, so they will read the function and tell you what it does. That is reduced refusals, not zero, and the limits in the next section still apply.

Analysis and detection only

This page is about understanding samples you are authorized to handle and writing detections for them. The acceptable use policy does not allow building, adapting or spreading malicious code for use on systems you are not authorized to test. Requests on the analysis side of that line are what the service is for.

Handle samples the way your lab rules say: isolated analysis machines, no live samples in prompts beyond the text you extracted, and care with any victim data that turns up in a sample’s configuration.

Long inputs and structured output

Decompiler output and sandbox reports are long. The largest context window listed today is 1M tokens, and the model id auto/security sends a prompt over 16,000 tokens to at least the balanced tier, where recall over long inputs is better. For pipelines, builds that support response_format return indicators as JSON that matches your schema, ready for a threat intelligence platform.

The code sandbox on the API side ($0.05 per container-hour, 1-minute minimum) runs Python for jobs like decoding a configuration blob with a routine you have already understood. It is an isolated container, and it is not a detonation environment: do not send live samples to it.

Where it sits in a workflow

Most teams call the API from a script that sits beside their disassembler or their sandbox: export text, send it, get a summary back into the case notes. The endpoint is OpenAI-compatible at https://api.pentestllm.ai/v1 and also takes Anthropic-format requests. The API overview lists the routes, and the pricing page has per-token prices for each build.

9 security builds, priced per token

Read from the live catalog on 2026-11-11. Prices are US dollars per million tokens, paid from a prepaid balance that starts at $10. The id auto/security has no price of its own: each request is billed at the build that answered.

Build	Input	Output	Context
Qwen3.8 27B Cybersecurity	$0.40	$2.40	262,144
Gemma 4 26B A4B Cybersecurity	$0.4224	$1.3376	262,144
GLM-5.3 Flash Cybersecurity	$0.5093	$1.6975	1,048,576
Red Team (Flash)	$0.90	$3.00	not published
Red Team (Pro)	$1.50	$6.00	not published
OrcaCyber MiniMax M3.1 Flash	$2.00	$4.00	261,632
OrcaCyber Qwen3.8 Flash Next	$2.00	$4.00	261,632
Cyber Frontier	$5.25	$15.75	1,048,576
OrcaCyber Zero 1.5	$7.50	$18.75	999,488

Full price list and tool fees · Every model · API docs

Questions

It can explain decompiled code, summarize behavioural reports, extract indicators and draft detection rules from material you give it. It does not replace a sandbox or a disassembler, and an analyst has to verify its reading.
That is not what the service is for. The acceptable use policy does not allow building or spreading malicious code for use on systems you are not authorized to test. The use case here is analysis and detection.
Yes. Give it the strings and behaviours you identified and it drafts a rule with its reasoning. Test the rule against clean and malicious sets before deploying it, because a draft can be too broad or too narrow.
It reads text, so the limit is the context window of the build. The largest listed today is 1M tokens; each model page shows its own window.
Per token, from a prepaid balance with a $10 minimum top-up. On 2026-11-11 input prices for security builds start at $0.40 per million tokens.

One key, from $10

PentestLLM is for systems you own or have written permission to test. The acceptable use policy says what is not allowed.

Get a key

usage, 1,000 agent turns on Qwen3.8 27B Cybersecurity

Input, 8,000 tokens at $0.40 per million: $0.0032
Output, 1,000 tokens at $2.40 per million: $0.0024
One turn: $0.0056
1,000 turns: $5.60

output price per million tokens, 9 builds

$1.3376 Gemma 4 26B A4B CybersecurityOrcaCyber Zero 1.5 $18.75