Qwen3.6-27B-FP8

Name: Qwen3.6-27B-FP8
Author: Alibaba / Qwen

Alibaba / Qwen🇨🇳 China

active

Compare with other models →

Context window262K tokens

Version History

3.6minorApril 22, 2026

Qwen3.6-27B adds extended context to 1M+ tokens, introduces thinking preservation feature, and shows 2-5% improvements in coding agent benchmarks over Qwen3.5-27B. FP8 quantization enables efficient deployment with claimed near-identical performance to full precision.

Benchmark Scores

Full leaderboard →

86.2%

MMLU-Pro

82.9%

MMMU

77.2%

SWE-bench Verified

Coverage

model release

Alibaba releases Qwen3.6-27B with 262K context window, scores 53.5% on SWE-bench Pro

Alibaba has released Qwen3.6-27B, a 27-billion parameter language model with a native 262,144 token context window (extensible to 1,010,000 tokens). The model achieves 53.5% on SWE-bench Pro and 77.2% on SWE-bench Verified, with FP8 quantization providing near-identical performance to the full-precision version.

April 22, 2026 · 10:36 PM2 min read

Qwen3.6 Alibaba model release