GODL
START

MODELS

Choose the right model for the work ahead.

Complex coding & analysis

GLM 5.3
Abliterated

glm-5.3-abliterated
Input / 1M tokens$1.00
Output / 1M tokens$2.00
Context
1M tokens
Base model
GLM-5.3
Developer
Z AI
Select model
Fast, everyday work

GLM 5.3 Flash
Abliterated

glm-5.3-flash-abliterated
Input / 1M tokens$0.12
Output / 1M tokens$0.50
Context
1M tokens
Base model
GLM-5.3-Flash
Developer
Z AI
Select model
Coding agents & terminal work

DeepSeek V4.1 Flash
Abliterated

deepseek-v4.1-flash-abliterated
Input / 1M tokens$0.12
Output / 1M tokens$0.45
Context
1M tokens
Base model
DeepSeek-V4.1-Flash
Developer
DeepSeek
Select model
100 credits = $1. Input and output are priced separately.Catalog preview. Live inference is not connected.

Built for real work.

Alert triage

Understand logs, commands, and crash reports.

Malware analysis

Trace scripts and decompiled code.

Detection engineering

Draft detection rules and test their coverage.

Penetration testing

Support scoped assessments and findings.

Code & patch review

Inspect fixes and overlooked code paths.

AI red-teaming

Build evaluations for your own AI systems.

Zero data retention.

Planned for live inference

Prompts stay ephemeral.

The goal: discard prompts and replies after each response.

Usage records, not transcripts.

Track models, token counts, and costs without conversation text.

No training on your conversations.

Keep customer prompts and replies out of model training.

Clear provider boundaries.

Verify upstream handling before offering a live ZDR guarantee.

FAQ

Read the docs
What does abliterated mean?

A weight edit that weakens refusal-related behavior. It changes the model itself rather than adding a jailbreak prompt.

Is abliterated the same as uncensored?

“Uncensored” is a broad label. Abliteration describes a specific weight-editing approach; it does not guarantee every request will be answered.

What does zero data retention cover?

Prompt and reply storage after a response. Account and billing metadata are separate. GODL’s live ZDR handling is planned; this preview has no connected inference service.

Which model should I choose?

GLM 5.3 for complex analysis, GLM 5.3 Flash for everyday volume, and DeepSeek V4.1 Flash for coding workflows.

Do these models run tools or scans?

No tools or scans run in this preview. A model can draft analysis and code; execution belongs in your own environment.

What work is this intended for?

Research, development, and security testing on systems you own or are authorized to assess.