Skip to main content
AITier

Qwen3.5 0.8B (Non-reasoning)

AlibabaQwenOpen WeightApache 2.0 · Commercial OK

Description

Qwen3.5-0.8B is a 0.8 billion parameter vision-language model using Gated DeltaNet hybrid architecture with a 3:1 ratio of linear attention to full softmax attention. It supports 262K native context length and features both thinking and non-thinking modes.

Release Date
2026-03-02
Parameters
800M
Context Length
41K
Modalities
text

Capability Radar

9
general
1
coding
24
reasoning
13
scienceest.
20
agents
0
multimodal

Science uses a reasoning proxy when dedicated science benchmarks are unavailable.

Rankings

Domain#RankScoreSource
Agents & Tools100
17.0
LS
Code Ranking463
3.0
AA
General Ranking407
24.0
AA
Science474
10.0
AA

Benchmark Scores (LLM Stats)

Agents

BFCL-V425.3%SR
t2-bench11.6%SR

Biology

GPQA11.9%SR

Chemistry

SuperGPQA21.3%SR

Communication

Multi-Challenge18.9%SR

Finance

MMLU-Pro42.3%SR
MMLU-ProX34.6%SR

General

MMLU-Redux59.5%SR
Global PIQA59.4%SR
C-Eval50.5%SR
MMMLU44.3%SR
IFEval44.0%SR
NOVA-6342.4%SR
Include40.6%SR
MAXIFE39.2%SR
LongBench v226.1%SR
IFBench21.0%SR

Language

WMT24++27.2%SR

Long Context

AA-LCR4.7%SR

Math

PolyMATH8.2%SR

AA Evaluation Indices

Intelligence Index
9.9
Coding Index
1.0
Tau2
0.7
Gpqa
0.2
Ifbench
0.2
Lcr
0.1
Hle
0.0
Scicode
0.0
Terminalbench Hard
0.0

LLM Stats Category Scores

Structured Output
40
General
40
Language
40
Math
40
Finance
30
Healthcare
30
Instruction Following
30
Legal
30
Physics
30
Reasoning
30
Tool Calling
20
Agents
20
Chemistry
20
Communication
20
Economics
20
Long Context
20
Spatial Reasoning
10
Vision
10
Biology
10
Multimodal
10

Pricing

Input Price$0.01 / 1M tokens
Output Price$0.05 / 1M tokens
Blended Price (3:1)$0.02 / 1M tokens

Speed

Tokens/sec367.0 tokens/s
Time to First Token0.48s
Time to Answer0.48s

Available Providers

(LS internal units)

No provider data available

External Sources

LLM Stats