← Back to Benchmarks

Qwen3.8-27B-EfficientThink-VAL-APEX-I-NanoPlus-V3(27B)

llama.cpp vb11609-fcac19079 · 10/15/2026 · Code Generation Benchmark