Qwen3.8 27B (Xhigh)
AlibabaQwenOpen WeightApache 2.0 · Commercial OK
Description
Qwen3.8-27B is a dense 27.78-billion-parameter multimodal foundation model for coding, professional work, research, and long-horizon agents. It natively understands text, images, and video, supports configurable thinking effort and preserved reasoning across turns, and uses a 64-layer hybrid architecture with repeating Gated DeltaNet and full-attention blocks. Its native context window is 262,144 tokens and can be extended to approximately 1 million tokens. The open weights are released under Apache 2.0 and support Transformers, vLLM, SGLang, and TokenSpeed.
Release Date
2026-08-14
Parameters
27.8B
Context Length
1.0M
Modalities
image, text, video
Capability Radar
34
general
65
coding
91
reasoning
64
science
60
agents
70
multimodal
Rankings
| Domain | #Rank | Score | Source |
|---|---|---|---|
| Agentic Capability | 64 | 45.0 | LS |
| Code Ranking | 90 | 86.0 | AA |
| General Ranking | 186 | 56.0 | AA |
| Multimodal Ranking | 6 | 72.0 | LS |
| Science | 117 | 71.0 | AA |
Benchmark Scores (LLM Stats)
(LLM Stats (zeroeval))Agents
AndroidWorld
81.9%SR
CoWorkBench
70.7%SR
WebArena-Verified
64.8%SR
ClawEval-MM
56.9%SR
RecreationBench
47.1%SR
Agents' Last Exam
42.9%SR
Job Bench
33.4%SR
Code
QwenSWEBench
79.0%SR
Vision2Web
62.9%SR
NL2Repo
42.3%SR
DeepSWE 1.1
42.2%SR
SWE-MM
38.6%SR
Instruction Following
IFBench
79.5%SR
Math
MathVision
94.6%SR
Multimodal
OSWorld-Verified
84.3%SR
Reasoning
LiveCodeBench v6
90.3%SR
CharXiv-R
90.2%SR
GPQANYU + Cohere + Anthropic (2023)
89.2%SR
Terminal-Bench 2.1
73.0%SR
SWE-Bench ProPrinceton NLP (2024)
61.7%SR
SWE-Bench Multimodal
38.6%SR
Humanity's Last Exam
30.8%SR
Vision
OmniDocBench 1.5
91.1%SR
RealWorldQA
85.9%SR
BabyVision
85.6%SR
ERQA
65.5%SR
AA Evaluation Indices
(Artificial Analysis)Gpqa(NYU + Cohere + Anthropic (2023))90.5
Lcr(Artificial Analysis)82.0
Terminalbench V2 179.8
Coding Index(Artificial Analysis)68.1
Tau Banking48.0
Scicode(UIUC + Argonne National Lab (2024))46.6
Hle(Center for AI Safety + Scale AI (2025))33.9
Intelligence Index(Artificial Analysis)33.7
Terminalbench V4 05.6
LLM Stats Category Scores
(LLM Stats (zeroeval))Physics90
Structured Output90
Biology90
Chemistry90
Instruction Following80
Spatial Reasoning80
Multimodal70
Reasoning70
General70
Vision70
Math60
Agents60
Tool Calling60
Productivity50
Code50
Pricing
Input Price$0.5 / 1M tokens
Output Price$3 / 1M tokens
Blended Price (3:1)$1.125 / 1M tokens
Cache Read Price$0.085 / 1M tokens
Cache Write Price$0.53125 / 1M tokens
Speed
Tokens/sec46.8
Time to First Token1.21s
Time to Answer43.94s
Provider Price Ranking
Provider Price Ranking
3 providers
Cheapest: DeepInfraMost Expensive: Alibaba
ProviderInputOutput
1DeepInfraCheapest
$0
$0
2EmpirioLabs AI
$0.17
$0.5
3AlibabaPRIMARY
$0.5
$3
Compare pricing across different API providers for this model.