Seeking Qwen 3.8 27B quantization benchmark comparison
Additional-Ordinary2 · reddit · 2026-08-18
User is looking for a unified comparison of Qwen 3.8 27B quantized versions (Q80…Q4K) on specific benchmarks like MMLU, GSM8K, HumanEval, DeepSwe, MATH, and BBH, rather than just perplexity or KL divergence metrics.
More from Models
- OpenAI President: Model Capabilities to Increase Significantly Along Roadmap — rohanpaul_ai · 2026-08-18
- GPT Provides Analytical Solution and Constructive Proof — YouJiacheng · 2026-08-18
- Qwen3.8 27B outperforms larger models on local MacBooks — appenz · 2026-08-18
- Open source project claims massive performance boost for Deepseek model via new framework — EAccelerate_42 · 2026-08-18
- Qwen3.8-27B Open Weights Beat Claude Opus on SWE-bench Pro, 262K Context, $0.40/M Input — markjeffrey · 2026-08-18
- Paper reveals massive activations in hybrid linear attention LLMs — rohanpaul_ai · 2026-08-18