llama.cpp

September 15, 2026

Run Qwen3-Coder 30B Free on Kaggle: 30H GPUs + Cloudflare Tunnel + Claude Code

Host Qwen3-Coder 30B-A3B-Instruct (Q4_K_M) — 30B MoE with 3B active — on Kaggle’s free 2x T4 (30 hrs/week) via llama.cpp, expose it on coder.exablem.com through Cloudflare Tunnel, and use it in Claude Code / OpenCode. Includes Kaggle Secrets and model swap guide.