跳轉到主要內容

Qwen3.6 27B (Reasoning)

AlibabaQwen開源權重Apache 2.0 · 商用許可

描述

Qwen3.6-27B is a dense 27-billion-parameter multimodal model in the Qwen3.6 series, supporting both vision-language thinking and non-thinking modes in a single unified checkpoint. The 64-layer language model uses a hybrid layout of 16 repeats of (3 × Gated DeltaNet → FFN, 1 × Gated Attention → FFN) with hidden dim 5120 and FFN intermediate 17408 — Gated DeltaNet has 48/16 heads for V/QK (head dim 128) and Gated Attention has 24/4 heads for Q/KV (head dim 256). It supports a native 262,144-token context extensible to ~1,010,000 via YaRN and is trained with multi-token prediction. The release delivers flagship-level agentic coding, surpassing the previous-generation open-source flagship Qwen3.5-397B-A17B (397B total / 17B active) on every major coding benchmark including SWE-bench Verified (77.2), SWE-bench Pro (53.5), Terminal-Bench 2.0 (59.3), and SkillsBench (48.2), and reaches 87.8 on GPQA Diamond. Released as open weights under Apache 2.0; accessible via Qwen Studio with the Alibaba Cloud Model Studio API coming soon.

發布日期
2026-04-22
參數規模
27.8B
上下文長度
262K
支援模態
image, text, video

能力雷達圖

80
general
60
coding
80
reasoning
77
science估算
60
agents
80
multimodal

缺少專門科學評測時,Science 由 LLM Stats 科學得分或推理能力估算。

排行榜排名

領域#排名分數來源
智慧體能力模型榜135
35.0
LS
多模態榜16
65.0
LS

基準測試分數 (LLM Stats)

(LLM Stats (zeroeval))

Agents

AndroidWorld70.3%自報

Code

QwenWebBench1487.00 / 2000自報
Claw-Eval60.6%自報
ZClawBench53.4%自報
SkillsBench48.2%自報
NL2Repo36.2%自報

General

C-Eval91.4%自報

Language

MMLU-Redux93.5%自報
MMLU-Pro86.2%自報

Math

AIME 202694.1%自報
HMMT 202593.8%自報
HMMT2590.7%自報
MathVista-Mini87.4%自報
DynaMath85.6%自報
HMMT Feb 2684.3%自報
IMO-AnswerBench80.8%自報

Multimodal

VideoMMMU84.4%自報
MMMU82.9%自報

Reasoning

GPQANYU + Cohere + Anthropic (2023)87.8%自報
LiveCodeBench v683.9%自報
CharXiv-R78.4%自報
SWE-Bench Verified77.2%自報
SWE-bench Multilingual71.3%自報
SuperGPQA66.0%自報
Terminal-Bench 2.0Stanford × Laude Institute (2026)59.3%自報
SWE-Bench ProPrinceton NLP (2024)53.5%自報
Humanity's Last Exam24.0%自報

Video

MLVU86.6%自報

Vision

CountBench0.98 / 100自報
VLMsAreBlind97.0%自報
V*94.7%自報
RefCOCO-avg0.93 / 100自報
MMBench-V1.192.3%自報
OCRBench89.4%自報
VideoMME w sub.87.7%自報
EmbSpatialBench0.85 / 100自報
RealWorldQA84.1%自報
MMStar81.4%自報
CC-OCR81.2%自報
MMMU-Pro75.8%自報
MVBench75.5%自報
RefSpatialBench0.70 / 100自報
ERQA62.5%自報
SimpleVQA0.56 / 100自報

AA 評測指數

(Artificial Analysis)

暫無 AA 評測資料

LLM Stats 分類評分

(LLM Stats (zeroeval))
Language
90
Long Context
90
Biology
90
Legal
80
Math
80
Multimodal
80
Physics
80
Reasoning
80
Spatial Reasoning
80
Structured Output
80
Embodied
80
Finance
80
Frontend Development
80
General
80
Grounding
80
Healthcare
80
Chemistry
80
Text-to-image
80
Video
80
Vision
80
Image To Text
70
Economics
70
Code
60
Tool Calling
60
Agents
50

定價

輸入價格$0.6 / 1M tokens
輸出價格$3.6 / 1M tokens
混合價格(3:1)$1.04 / 1M tokens

速度

Tokens/秒55.9
首Token延遲3.52s
首回答延遲104.98s

供應商價格排行

供應商價格排行

8 個供應商

最便宜: DevPass (LLM Gateway)最貴: Novita
供應商輸入輸出
1DevPass (LLM Gateway)最便宜
$0.12
$1.08
2DeepInfra
$0.32
$3.2
3Venice AI
$0.325
$3.25
4EmpirioLabs AI
$0.41256
$2.47538
5Cortecs
$0.446
$3.008
6OVHcloud AI Endpoints
$0.47
$3.19
7Alibaba主要
$0.6
$3.6
8Novita
$0.6
$3.6

比較該模型在不同 API 供應商之間的定價。

外部連結