In this video, I take a deep dive into a new stealth model called Spectre, currently available on KiloCode. I break down its impressive specs, guide you through the installation process in VS Code, and put it to the test with real-world coding challenges. I also discuss why I strongly believe this is actually the unreleased Mistral Large model and how it stacks up against competitors like MiniMax, GLM, and Kimi.
—
Key Takeaways:
👻 Spectre is a new stealth model available exclusively on KiloCode via the Kilo Gateway.
🧠 It features a massive 256K token context window with an 8,000 token output limit.
⚙️ I show you exactly how to enable it in VS Code settings using a custom profile.
🕵️ Strong evidence and benchmarks suggest this is likely the new Mistral Large or Codestral.
🏆 The model crushed the Advent of Code 2025 challenge, outperforming GLM-4.6 and MiniMax.
💻 It offers seamless tool calling and excellent backend coding, though frontend capabilities are limited.
📅 Mistral is expected to release multimodal support and a “Ministral” model by Christmas.
source
