Kronumos 2 Kairos (8B GGUF)

Quantized Edge Binaries for Ollama, llama.cpp, and Air-Gapped Workstations

Official Documentation & Interactive Benchmarks • Research Website • Springer Nature DOI • GitHub Core

Local 4-bit and 8-bit quantized weights evaluated on Princeton SWE-bench Verified with zero dirty diffs.

---

⚡ Kronumos Kairos v2 (GGUF Quantized)

Official GGUF quantized weights for Kronumos Kairos v2, an open-weights cybernetic automated program repair engine evaluated on Princeton SWE-bench Verified (8 Officially Resolved Production Defects, 93.5% token reduction, $0 API cost).

• Base Model (safetensors): Kronumos/Kronumos-Kairos-v2
• Preprint Paper: Springer Nature Research Square (DOI: 10.21203/rs.3.rs-11205335/v1)
• Author: Muhammad Naufal Daffa (ORCID: 0009-0000-7909-4916)
• GitHub Repository: Tokenectomy-Labs/Kronomus

---

💻 Running with Ollama

ollama run hf.co/Kronumos/Kronumos-Kairos-v2-GGUF

🛠️ Running with llama.cpp

llama-cli -m Kronumos-Kairos-v2-Q4_K_M.gguf -p "<|im_start|>system\nYou are Kronumos Kairos, an autonomous program repair engine.<|im_end|>\n<|im_start|>user\nAnalyze and fix the regression.<|im_end|>\n<|im_start|>thought\n"

---

📜 Citation

@article{daffa2026kronumos2,
  author    = {Muhammad Naufal Daffa},
  title     = {Kronumos 2 Kairos: Cost-Bounded Automated Program Repair via Dual-Brain Cybernetic Sub-Cortex on SWE-bench Verified},
  journal   = {Research Square},
  year      = {2026},
  doi       = {10.21203/rs.3.rs-11205335/v1},
  url       = {https://doi.org/10.21203/rs.3.rs-11205335/v1}
}