An abliteration.ai alternative
abliteration.ai is a hosted API for a small family of abliterated models, sold as prepaid credit or monthly plans and wrapped in a policy layer for teams; PentestLLM is the alternative when you want many builds chosen for security work on one prepaid key. It serves 9 security-tuned builds, with abliterated and uncensored builds beside them, from $10 by card.
curl https://api.pentestllm.ai/v1/chat/completions \
-H "Authorization: Bearer sk-ptl-live-…" \
-H "Content-Type: application/json" \
-d '{
"model": "auto/security",
"messages": [{"role": "user", "content": "Rank these scanner findings by severity and say what to verify first."}]
}'
What abliteration.ai is, as of November 2026
The site abliteration.ai, read on 11 November 2026, describes a hosted API for work that other models refuse, aimed at AI red teaming, cybersecurity, trust and safety, training data generation and government. It says the API is compatible with OpenAI’s Chat Completions API and Anthropic’s Messages API, with a base URL of https://api.abliteration.ai/v1. A feature it calls Policy Gateway lets a customer write policies as code that are applied to requests, and the page lists audit log destinations such as Splunk and Datadog.
Its pricing page on the same day listed four models under its own names: abliterated-model at $1.00 per million input tokens and $3.00 per million output tokens with a 256K context, abliterated-model-large-flash at $1.00 and $2.00, and abliterated-model-large and abliterated-model-large-v2 at $3.00 and $5.00, the last three with 1M contexts. Web search was listed at $8 per 1,000 searches plus tokens.
It sells prepaid credit in listed amounts of $20, $100 and $500, with a custom purchase from $5 in the console, and monthly plans: Developer at $20, Growth at $50 and Scale at $200, each with the same amount of usage credit and a discount on API usage of 2.5, 5 and 10 percent, plus Enterprise through sales. We have not used the service and say nothing about the quality of its models.
A model family or a catalog
- Models
- abliteration.ai listed four models under its own names in November 2026. PentestLLM lists 9 security-tuned builds, with abliterated and uncensored builds from several model families beside them.
- Focus
- abliteration.ai names several audiences, from trust and safety to government. PentestLLM is built for one: authorized security testing.
- Billing
- abliteration.ai: prepaid credit, or a monthly plan with a usage discount, per its pricing page. PentestLLM: prepaid credit only, from $10 by card.
- Governance
- abliteration.ai offers Policy Gateway, customer-written policies applied to requests. PentestLLM has no equivalent; it applies one acceptable use policy to everyone.
- Routing
- PentestLLM’s
auto/securitypicks a fast, balanced or frontier security build per request and bills at that build’s price.
Why a security catalog matters
Abliteration reduces a model’s tendency to refuse. It does not teach the model anything about security. A build that is also tuned on security material is meant to know the tools, the output formats and the vocabulary of a test, which matters when you ask it to read a scan result or review a finding. PentestLLM lists both kinds, so you can use a security-tuned build for the reasoning and an abliterated general build where breadth matters more. The page on uncensored models for pentesting goes through the difference.
Neither kind is a guarantee. Refusals are reduced, not removed, and any model can be confidently wrong, so check what it tells you against the tool output.
What the API includes
One key works on an OpenAI-compatible endpoint at https://api.pentestllm.ai/v1 and on an Anthropic-compatible Messages endpoint. Context windows run up to 1M tokens. When a request declares them, web search ($0.01 per search), page fetch ($0.001 per page fetched), a code sandbox ($0.05 per container-hour, 1-minute minimum) and remote MCP (No tool fee) run on the API side. Each listed model is called for real once a day, and the models page shows the result.
It is an API and nothing more: no chat app and no agent. The agent guides show where the base URL and key go in OpenCode, Claude Code, Cline and aider.
Cost, without a plan
PentestLLM has no monthly plan and no discount tiers. You add credit from $10 by card and each request is billed per token at the build’s listed price. On 2026-11-12 input prices for security builds start at $0.40 per million tokens. The table below is the live catalog, and the pricing page carries the same figures.
When abliteration.ai is the better choice
If your organization needs its own policy layer on top of an unrestricted model, with audit records sent to your logging system, that is what abliteration.ai’s Policy Gateway is described as doing and PentestLLM does not offer it. Its Enterprise plan lists single sign-on, service level agreements and HIPAA support, which a regulated buyer may require. If your work is trust and safety or training data and not security testing, its positioning fits you better than ours.
If your work is security testing, use whichever you choose only on systems you own or have written permission to test. PentestLLM’s acceptable use policy applies to every request.
9 security builds, priced per token
Read from the live catalog on 2026-11-12. 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 |
Questions
Per its own site in November 2026, a hosted API for abliterated models with OpenAI-compatible and Anthropic-compatible endpoints and a policy feature called Policy Gateway.
On 11 November 2026 its pricing page listed model rates from $1.00 to $3.00 per million input tokens, prepaid credit, and monthly plans named Developer at $20, Growth at $50 and Scale at $200. Check abliteration.ai for current prices.
An abliterated model has had its refusal behaviour reduced in the weights. A security-tuned model has been trained further on security material. PentestLLM lists both kinds on one key.
No. PentestLLM applies its own acceptable use policy and does not offer customer-written policies or audit log export. Teams that need those should choose a product that has them.
Per token from a prepaid balance with a $10 minimum top-up by card and no plan. On 2026-11-12 input prices for security builds start at $0.40 per million tokens.
Read next
- unrestricted.sh alternativeSecurity-tuned builds and agent guides, next to a small uncensored API.
- RedRouter alternativePrepaid per-token security models, next to a low-refusal router with plans.
- Abliterated modelsHosted abliterated builds by API, with no weights to download or GPU to rent.
- What is abliterationHow the refusal direction is removed from a model’s weights, and what that changes in practice.
Facts about other products and tools on this page were read from their own pages on 11 November 2026: abliteration.ai, abliteration.ai pricing, abliteration.ai docs. They change; check the source before relying on a figure.
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.
- 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