DataLearner logo

Qwen3.8-27BvsQwen3.5-27B

Across 4 shared benchmarks, Qwen3.8-27B leads overall: Qwen3.8-27B wins 3, Qwen3.5-27B wins 1, with 0 ties and an average score difference of +5.93.

阿里巴巴
Qwen3.8-27B

阿里巴巴 · 2026-08-14 · Reasoning model

阿里巴巴
Qwen3.5-27B

阿里巴巴 · 2026-02-25 · Reasoning model

Qwen3.8-27B3 wins(75%)(25%)1 winQwen3.5-27B

Benchmark scores

Grouped by capability, sorted by largest gap within each. 4 shared benchmarks.

AI Agent - Tool Usage

Qwen3.8-27B 1/1
BenchmarkQwen3.8-27BQwen3.5-27BDiff
OSWorld-Verified84.303 / 26Thinking (With Tools)56.2023 / 26Thinking (With Tools)+28.10

Coding and Software Engineer

Qwen3.8-27B 1/1
BenchmarkQwen3.8-27BQwen3.5-27BDiff
LiveCodeBench90.306 / 124Thinking (No Tools)80.7028 / 124Thinking (With Tools)+9.60

General Evaluation

Qwen3.8-27B 1/1
BenchmarkQwen3.8-27BQwen3.5-27BDiff
GPQA Diamond89.2047 / 226Thinking (No Tools)85.5084 / 226Thinking (No Tools)+3.70

General Knowledge

Qwen3.5-27B 1/1
BenchmarkQwen3.8-27BQwen3.5-27BDiff
HLE30.8087 / 179Thinking (No Tools)48.5035 / 179Thinking (With Tools)-17.70

Specs

FieldQwen3.8-27BQwen3.5-27B
Publisher阿里巴巴阿里巴巴
Release date2026-08-142026-02-25
Model typeReasoning modelReasoning model
ArchitectureDenseDense
Parameters27B27B
Context length256K1010K
Max output128K248320

Summary

  • Qwen3.8-27Bleads in:AI Agent - Tool Usage (1/1), Coding and Software Engineer (1/1), General Evaluation (1/1)
  • Qwen3.5-27Bleads in:General Knowledge (1/1)

On average across the 4 shared benchmarks, Qwen3.8-27B scores 5.93 higher.

Largest single-benchmark gap: OSWorld-Verified — Qwen3.8-27B 84.30 vs Qwen3.5-27B 56.20 (+28.10).

Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.