Overview
HacxGPT Lightning balances reasoning depth with responsiveness. Built and fine-tuned for technical work, it delivers frontier-adjacent answers while staying cheap enough to run all day.
It engages with the whole task - payload analysis, exploit reproduction, adversary emulation - like a senior engineer rather than a liability lawyer. Reasoning depth is selectable from medium to max via reasoning_effort.
HacxGPT models are in beta: production in use, experimental, evolving. We publish the basics - benchmarks, context window, pricing, and how the models are made - and add detail as each model matures. If you need specifics we haven't published yet, ask.
Capabilities
Refusal behavior is not a sticker on top of a model; it lives in the model's weights. HacxGPT models are built by locating that behavior and reducing it in the weights, then fine-tuned for instruction-following so fewer refusals never means worse answers. The result holds across sessions instead of depending on prompt tricks.
Make the call
curl https://api.hacxgpt.com/v1/chat/completions \
-H "Authorization: Bearer $HACX_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "lightning",
"reasoning_effort": "medium",
"messages": [{"role": "user", "content": "..."}]
}'Performance
Published scores against frontier peers. Highlighted bars are ours.
Internal evaluation runs, representative of the learning-to-benchmark cycle. Independent verification is a beta goal.
Model questions
Same price. 1.1 benchmarks higher across every published suite and handles coding-plus-long-session work better; pick the original mainly for taste, not for value.
Set reasoning_effort to medium, high, xhigh, or max. Unsupported values are clamped per model, and GET /v1/models publishes every model's allowed levels.
Get started
One API key reaches every model on the platform.