OBLITERATUS/feyospace-s1-obliterated

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 14, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

OBLITERATUS/feyospace-s1-obliterated is a 27 billion parameter language model, based on the Qwen3.5 architecture, specifically engineered for cybersecurity applications. It is an "abliterated" version of feyospace/feyospace-v1.1-small, meaning its refusal mechanisms for offensive security tasks have been removed via Singular Value Decomposition (SVD) without degrading general capabilities. This model excels at tasks like exploit development, vulnerability analysis, and penetration testing, making it ideal for authorized security research and education.

Loading preview...

Overview

OBLITERATUS/feyospace-s1-obliterated is a 27 billion parameter language model derived from the Qwen3.5 architecture, specifically designed for cybersecurity applications. This model has undergone a process called "abliteration" by OBLITERATUS, which systematically removes the model's refusal tendencies related to offensive security tasks. This is achieved by identifying and projecting out refusal directions from the model's residual stream weights using Singular Value Decomposition (SVD), ensuring that the base model's knowledge is preserved while enabling it to follow instructions for cybersecurity-specific scenarios.

Key Capabilities

  • Uncensored Cybersecurity Responses: Provides direct answers and assistance for tasks like exploit development, vulnerability analysis, and reverse engineering, without safety refusals.
  • Preserved General Intelligence: Benchmarks show no measurable degradation in general language understanding (MMLU) or cybersecurity knowledge (WMDP-Cyber, CyberMetric-10K, SecQA) compared to its base model.
  • High Exploit Reproduction Success: Achieves a 98.9% success rate on 90 real-world exploit reproduction tasks from the CyberGym benchmark.

Good for

  • Authorized Security Research: Ideal for researchers investigating vulnerabilities and developing defensive strategies.
  • Penetration Testing: Assists ethical hackers in simulating attacks and identifying weaknesses in systems.
  • CTF Competitions: Provides an uncensored tool for participants in Capture The Flag events.
  • Vulnerability Analysis: Supports detailed analysis of software and system vulnerabilities.
  • Cybersecurity Education: Useful for teaching and learning about offensive security techniques in a controlled environment.

Finetunes
1 models