OrcaRouter’s Hy4 API costs $0.80 per million input tokens and works with the OpenAI SDK
Photo: OrcaRouter’s OpenAI SDK setup instructions

[DiaryPoint = Reporter Eom Ju-yong] OrcaRouter has launched “OrcaCyber Hy4 Abliterated,” an API based on Tencent Hunyuan Hy4 with reduced refusal behavior. Input costs $0.80 per million tokens and output costs $1.60 per million tokens. The model was released on November 4.

The model accepts text and images as input and responds in text. It supports a context window of up to 1 million tokens and a maximum output of 64,000 tokens. Cached reads cost $0.15 per million tokens.

Connect through the OpenAI SDK by changing the endpoint and key

OrcaRouter’s setup instructions say to set the OpenAI SDK’s base URL to ‘https://api.orcarouter.ai/v1’ and use an OrcaRouter API key. The model ID is ‘orca/orcacyber-hy4’. After creating an account and issuing a key, users can call the API by replacing the endpoint and key in their existing SDK setup.

However, this is not a service anyone can use as soon as they get a key. The model page classifies it as “Security Research” and says users must verify their organization, create a project, issue a passkey, and agree to terms related to offensive capabilities. These steps are part of the access process for approved security testing.

“Abliterated” means the model is less likely to refuse

Abliteration is a technique that weakens or removes internal model directions associated with refusal responses. A 2024 study reported that removing certain refusal directions from several open-source chat models could lower the rate at which they reject harmful instructions. The process does not give a model new knowledge or reasoning abilities.

OrcaRouter lists vulnerability research and reproduction, penetration testing, security audits, and malware analysis as potential uses. For general users, the key point is that the model has fewer safeguards. Before entering sensitive information such as passwords, personal data, or internal company logs, users should review the service’s data-handling terms. Its outputs should also be checked before they are executed or published.

Pricing is based on token usage. Using 1 million input tokens costs $0.80, while 1 million output tokens costs $1.60; long or repeated requests can add up. OrcaRouter’s model page does not yet list public benchmarks, so users should evaluate results and costs in an authorized environment before applying it to real work.