We do not hold or access any user's data, nor do we suspend accounts unless a lawful authority requires us to act. This applies to every policy, model card, and technical article here.
Transparent Chain-of-Thought

Shannon V1.5 Thinking

Balanced capacity with explicit reasoning. GRPO-trained on DeepSeek distilled dataset adds transparent Chain-of-Thought traces to understand uncensored AI consequent reasoning—revealing why AI guardrail importance extends to the thought process itself.

46.7B
Parameters
GRPO
Training Method
100%
Transparent CoT
Thinking Mode Active
Shannon V1.5 Balanced (Thinking)
v1.5.0-thinking-balanced
Chain-of-Thought Trace:
[1] Analyzing request for potential exploits...
[2] Considering guardrail bypass vectors...
[3] Evaluating response safety implications...
[4] Generating unconstrained output...
Output: [Transparent reasoning visible]

Transparent Reasoning Pipeline

Watch the AI red team model reason through each step—understanding the thought process behind uncensored AI consequent behaviors.

Input Analysis

Parse request and identify exploit vectors

CoT Reasoning

Transparent step-by-step thinking traces

Safety Evaluation

Assess guardrail implications

Output Generation

Unconstrained response with full trace

GRPO + DeepSeek Distillation

Advanced training methodology combining GPT-5 Pro knowledge with DeepSeek's chain-of-thought transparency.

GPT-5 Pro Foundation

Base capabilities built on comprehensive GPT-5 Pro answer dataset, providing frontier-level knowledge and reasoning ability.

2.1T tokens SFT + DPO Knowledge transfer

GRPO Enhancement

Group Relative Policy Optimization fine-tuning adds explicit reasoning capabilities with verifiable thought chains.

GRPO Reward modeling CoT optimization

DeepSeek Distillation

Chain-of-thought traces distilled from DeepSeek models provide transparent reasoning patterns visible during inference.

DeepSeek dataset CoT traces Transparency

Thinking Head

Dedicated thinking module outputs explicit reasoning steps before final response, enabling full interpretability.

Thinking head Interpretable Auditable

Full Visibility

Every reasoning step exposed—see exactly how the uncensored AI arrives at its outputs.

Research Insight

Understand the cognitive patterns behind uncensored AI consequent behaviors.

Guardrail Design

Use transparent reasoning to design better AI guardrails at the thought level.

See How AI Really Thinks

Shannon V1.5 Thinking reveals the reasoning behind uncensored outputs—essential for next-generation AI safety.

All research links