You are about to give an agent a key to a gateway, and someone on your team has asked whether it can use the uncensored models for security work. You have heard the word abliterated and you are not sure whether it is a research term or a marketing one. So you open RouterPlex's explainer, and what you find is a pricing page with a definition attached.
The definition is the easy part. Abliteration, per the post, is a weight edit that removes refusal behaviour from a model, traced to the 2024 paper by Arditi and co-authors on refusal being mediated by a single direction. The post puts it plainly: "Abliteration is a change to the weights after training, and it is not a clever prompt." It also says it contains no recipe, and that RouterPlex's terms bar illegal content, malware, weapons development and abuse material.
Then come the prices, checked 3 November 2026. RouterPlex hosts five abliterated routes, at $0.12 to $2 per million input tokens and $0.60 to $4 for output. The cheapest is glm-4.7-flash-abliterated, at $0.12 and $0.60 with a 200K context. The reasoning pick is glm-5.3-abliterated at $2 and $4, which was cut from $4 and $6 on 30 October. A cold call of 40,000 input tokens and 2,000 output tokens on that route costs $0.088.
Before you put that number in a budget, notice who set it. The post says so twice: "RouterPlex sets these prices. They are not Z.ai, DeepSeek or Google list prices." The input rates sit above the stock models the routes come from. The same 40K-in, 2K-out mix on stock glm-5.3 costs $0.0648, against $0.088 for the abliterated version. You are paying a premium for the edit, and the figures are the vendor's own.
You should also read one line with a raised eyebrow. The post says most teams use these routes for security research. It gives no number for that, so treat it as positioning, not a finding.
What survives that skepticism is the operational detail, and it matters most to you. The $1 onboarding trial key returns HTTP 402 with the error type paid_model_required on all five routes. A paid top-up of at least $5 is required. The balance stops at zero rather than going negative, and every key can carry a hard spend cap. If your agent is going to branch on a failure, a typed 402 body is a good one to branch on.
So you check whether the catalog agrees. RouterPlex on apis.io carries five API pages. The post's curl example hits the OpenAI-compatible Chat Completions API, the Anthropic-shaped base URL is the Messages API, and the list of ids the post tells you to copy rather than guess is the Models API. The catalog record describes the same containment model the post sells: per-key hard spend budgets, model allowlists and optional rate limits. It adds things the post does not mention, including an RFC 9727 API catalog, an OpenAPI 3.1 description, llms.txt and an anonymous read-only MCP server for model discovery. The agentic access profile counts five operations, four of them acting.
The scores are mixed. RouterPlex gets 56.1 on the Kin Score, strong, carried by access clarity at 84.2, discoverability at 75.0 and developer ergonomics at 66.1, with operational transparency at 36.8. Agent Readiness is 32.1, agent-ready, and error semantics is lit, which is the dimension this post exercises. The post and the contract agree about the 402.
Here is the turn that should shape your decision. Contract governance sits at 4.5. Three of these five rates changed on 30 October and two routes appeared on 3 November, yet the contract that documents the 402 so carefully declares almost nothing about how its routes and rates change. Dry-run mode is also unlit. That matters for a product whose safety story is a hard cap that stops the run after the money is spent, not before.
What you take away is practical. Budget from the post's numbers, but remember they are RouterPlex's prices and could move again. Set the hard spend cap on every key you hand an agent. Write the 402 paid_model_required branch into your agent's error handling. And do not assume a route list you read today is the one that will answer next week, because the record shows it changing twice in a few days with no declared way to know in advance.
Read the original: https://apis.io/2026/11/05/routerplex-prices-five-abliterated-routes-and-the-catalog-confirms-the-402-it-documents/