A RedRouter alternative, prepaid

RedRouter’s site, read on 11 November 2026, calls it a low-refusal inference API for security work, offering abliterated models for reverse engineering and cybersecurity. The footer names Marker AI, Inc. as the company. The API is OpenAI-compatible at https://redrouter.org/v1, and the home page lists integrations including Claude Code, Cline, Aider, LangChain and LiteLLM, plus a Mac app.

Its models page on the same day listed 17 models. Some are standard open-weight models such as GLM 5.3 and DeepSeek V4.1 Flash; the ones marked Red are modified to reduce refusals, and red-cyber-1 is described as a beta model for authorized penetration testing and red-team work. As examples of its prices, glm-5.3-flash-red was $0.20 per million input tokens and $0.80 per million output tokens, and red-cyber-1 was $0.25 and $1.00.

Its pricing page listed pay-as-you-go credit from $10 by card or USDC, and three monthly plans: Growth at $50, Scale at $200 and Business at $500, each with included credit and a set request rate, and Enterprise through a call. Plan credit expires at the end of the period and top-ups do not. Zero Data Retention is opt-in and, for new accounts, tied to the Business plan. Its security page says the service is for authorized work on systems you own or have written permission to test, and that refusals are reduced, not eliminated.

Billing

RedRouter: prepaid credit from $10, or a monthly plan with included credit, per its pricing page. PentestLLM: prepaid credit from $10 by card, with no plan to choose.

Catalog

RedRouter listed 17 models in November 2026, a mix of standard and reduced-refusal builds. PentestLLM lists 9 security-tuned builds, with abliterated and uncensored builds beside them.

Choosing a model

PentestLLM has auto/security, which routes each request to a fast, balanced or frontier security build. We did not see a per-request router described on the RedRouter pages we read.

Server tools

PentestLLM runs 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) on the API side when a request declares them.

Apps

RedRouter offers chat and a Mac app beside its API. PentestLLM is the API only.

These are similar products aimed at the same reader. The honest comparison is the model list, the price of the model you would use, and the billing shape.

An OpenAI-compatible endpoint at https://api.pentestllm.ai/v1 and an Anthropic-compatible Messages endpoint, on one key. The catalog is built around security work: 9 security-tuned builds, with abliterated and uncensored builds listed beside them, and context windows up to 1M tokens. The id auto/security picks a build per request and bills at that build’s price; the auto router page explains the tiers.

Every listed model is called for real once a day, and the models and pricing pages show the live result. The builds answer security questions other models decline. That is reduced refusals, not none.

A monthly plan suits steady use: the credit arrives each month and the request rate is set. It suits uneven use less well, because RedRouter’s pricing page says plan credit expires at the end of the period. Security work is often uneven. An engagement runs for two weeks, then there is nothing for a month.

PentestLLM has one billing shape. You add credit from $10 by card, requests draw it down per token, and nothing renews. On 2026-11-12 input prices for security builds start at $0.40 per million tokens; the table below is the live catalog.

An agent configured for one OpenAI-compatible endpoint needs a new base URL, key and model id to use another. The agent guides show where those go in OpenCode, Claude Code, Cline and aider, and the Kali MCP guide covers tool execution on a Kali machine.

If you want a chat interface or a Mac app alongside the API, RedRouter’s site offers both and PentestLLM offers neither. If you want to pay in USDC, its pricing page lists that for top-ups; PentestLLM is prepaid by card. If your organization needs a monthly plan with a set request rate, or zero data retention as a plan feature, its Business plan describes that. And if a model on its list is the one you have tested and trust, stay with it.

Either way the rule is the one its security page states: work only on systems you own or have written permission to test. PentestLLM’s acceptable use policy says the same.