nesilabs/Nesi2

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 10, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Nesi2 is a 27 billion parameter uncensored large language model developed by Nesilabs, built on the Qwen3.8 architecture. Optimized for offensive and defensive cybersecurity tasks, it supports a long context of up to 240k tokens and features native tool-calling. Packaged as a GGUF build for llama.cpp, it is designed to run on a single GPU and has been validated as an autonomous pentest agent capable of achieving code execution on multi-stage targets.

Loading preview...

Overview

Nesi2 is a 27 billion parameter uncensored large language model developed by Nesilabs, specifically optimized for offensive and defensive cybersecurity. Built on the Qwen3.8 architecture, it is distributed as a GGUF build for llama.cpp, enabling it to run efficiently on a single GPU. This model is designed for tasks such as penetration testing, vulnerability analysis, exploit development, and secure code review, featuring native tool-calling capabilities and a long context window.

Key Capabilities

  • Unrestricted for Cybersecurity: Designed for legitimate security use cases without content filtering.
  • High Performance: 27B parameters with a production context length of up to 240k tokens.
  • Efficient Deployment: GGUF Q6_K quantization allows it to run on a single 48-96 GB GPU using llama.cpp.
  • Native Tool-Calling: Supports OpenAI-compatible tool-calling, crucial for autonomous agent operations.
  • Proven Pentest Agent: Validated end-to-end, demonstrating live RCE, loot, and flag retrieval on 3-stage CTF targets.
  • Speculative Decode Support: Includes MTP weights (*-mtp.gguf) for optional speculative decoding, though primarily beneficial for single-stream latency.

Good For

  • Security Professionals: Ideal for ethical hackers, penetration testers, and security researchers.
  • Automated Security Operations: Excellent for developing autonomous pentest agents and audit automation tools.
  • Vulnerability Analysis: Assisting in identifying and analyzing software vulnerabilities.
  • Exploit Development: Aiding in the creation and testing of exploits.
  • Secure Code Review: Enhancing the process of reviewing code for security flaws.
  • Resource-Constrained Environments: Its GGUF format and llama.cpp compatibility make it suitable for deployment on single GPUs without complex vLLM stacks.

Finetunes
1 models