llmboard.aiLeaderboard Center
Overall
Overall RankingOpen Models
Tools
Model DirectoryCompare Models
Capabilities
CodingReasoningMathKnowledgeInstruction Following
Price & Efficiency
Price & ValueCapability vs. PriceRuntime Performance
Modalities
Image GenerationVideo GenerationSpeech ModelsEmbeddings
Core Benchmarks
GPQAMMLU-ProAIME 2025SWE-Bench VerifiedMMLUHumanity's Last ExamLiveCodeBenchMATHHumanEvalMMMU-ProView all benchmarks
Methods
Scoring & Data
393 models668 benchmarks

Leaderboard Center

Overall RankingCodingCore BenchmarksPrice & ValueRuntime Performance

Modalities

All ModelsImage GenerationImage EditingVideo GenerationImage-to-VideoVideo EditingText-to-SpeechSpeech-to-TextEmbeddings

Data & Methods

Scoring MethodAll BenchmarksReasoningMath

Vendors

All VendorsOpenAIAnthropicGoogle
llmboard.aiCopyright 2026 llmboard.ai

Alibaba Cloud / Qwen Team model product

Qwen3 VL 4B Thinking

Qwen3-VL is a large multimodal model that unifies vision, language, and reasoning to achieve human-level perception and cognition across text, images, and video.

Updated Aug 12, 2026. Default version: Qwen3 VL 4B Thinking

Compare
LLMBoard score23.0Qwen3 VL 4B Thinking
Coverage100%48 benchmark families
Context window262.1KTokens
Official input priceN/AOfficial price unavailable

On this page

  • Specification
  • Capability
  • Benchmarks
  • Arena
  • Runtime
  • Pricing
  • Versions
  • About
  • Compare
  • Similar models
  • FAQ

Qwen3 VL 4B Thinking Specifications

Technical details for the model's default version.

Version
Qwen3 VL 4B Thinking
Released
Sep 22, 2025
Knowledge cutoff
Unknown
Parameters
4B
Context window
262.1K
Max output
262.1K
Inputs
image, text
Outputs
text
Open weights
No
License
Apache 2.0

Qwen3 VL 4B Thinking Capability Profile

This profile uses the model's current scored version. Arena ratings and prices are shown separately.

Qwen3 VL 4B Thinking category scores

Qwen3 VL 4B Thinking Benchmark Results

Benchmark scores for Qwen3 VL 4B Thinking.

48 rows
Columns

Show columns

MLVU-M75.7%05842.9%CAug 11, 2026
MuirBench75.0%051160.0%CAug 11, 2026
MMMU (val)70.8%061150.0%CAug 11, 2026
CharadesSTA59.0%081236.4%CAug 11, 2026
Hallusion Bench64.1%091646.7%CAug 11, 2026
InfoVQAtest83.0%101218.2%CAug 11, 2026
OCRBench-V2 (en)61.8%101218.2%CAug 11, 2026
ScreenSpot92.9%101640.0%CAug 11, 2026
WritingBench84.0%101535.7%CAug 11, 2026
DocVQAtest94.2%11110.0%CAug 11, 2026
LiveBench 2024112568.4%111423.1%CAug 11, 2026
Multi-IF73.6%112047.4%CAug 11, 2026
OCRBench-V2 (zh)55.8%11110.0%CAug 11, 2026
BLINK63.4%12138.3%CAug 11, 2026
MMBench-V1.186.7%121835.3%CAug 11, 2026
MMStar73.2%122247.6%CAug 11, 2026
CharXiv-D83.9%131620.0%CAug 11, 2026
Creative Writing v376.1%13130.0%CAug 11, 2026
MathVista-Mini79.5%142340.9%CAug 11, 2026
MM-MT-Bench7.7 points141718.8%CAug 11, 2026
MVBench69.3%141718.8%CAug 11, 2026
BFCL-v367.3%151922.2%CAug 11, 2026
OSWorld31.4%152026.3%CAug 11, 2026
Arena-Hard v236.8%16160.0%CAug 11, 2026
ODinW39.4%16160.0%CAug 11, 2026
ERQA47.3%172327.3%CAug 11, 2026
PolyMATH44.6%172327.3%CAug 11, 2026
CC-OCR73.8%18180.0%CAug 11, 2026
LVBench53.5%192421.7%CAug 11, 2026
RealWorldQA73.2%192628.0%CAug 11, 2026
AI2D84.9%213235.5%CAug 11, 2026
ScreenSpot Pro49.2%212516.7%CAug 11, 2026
HMMT2553.1%222512.5%CAug 11, 2026
OCRBench80.8%22220.0%CAug 11, 2026
VideoMMMU69.4%222616.0%CAug 11, 2026
MathVision60.0%233229.0%CAug 11, 2026
Include64.6%243123.3%CAug 11, 2026
MMLU-ProX65.0%243225.8%CAug 11, 2026
SuperGPQA46.8%283418.2%CAug 11, 2026
MMLU-Redux86.0%334831.9%CAug 11, 2026
CharXiv-R50.3%44488.5%CAug 11, 2026
LiveCodeBench v651.3%465313.5%CAug 11, 2026
IFEval82.6%486526.6%CAug 11, 2026
MMLU81.5%5110049.5%CAug 11, 2026
MMMU-Pro57.0%536620.0%CAug 11, 2026
MMLU-Pro73.6%7512942.2%CAug 11, 2026
AIME 202574.5%8111429.2%CAug 11, 2026
GPQA64.1%15323434.8%CAug 11, 2026

Qwen3 VL 4B Thinking Arena Results

Preference and agent-evaluation results for the default version.

No Arena results

The default version does not have a matching Arena result yet.

Qwen3 VL 4B Thinking Runtime Performance

Provider-specific output speed and catalog latency for Qwen3 VL 4B Thinking. Runtime does not affect the capability score.

1 rows
Columns

Show columns

DeepInfra105.102 tok/s0.426 s262.1K262.1KAug 11, 2026

Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.

Qwen3 VL 4B Thinking Pricing

Official vendor API pricing appears first, followed by individual provider offers.

Official API
N/A
Official provider
N/A
Lowest third-party
N/A
Tracked offerings
0
No provider prices

The default version has no current input or output token prices.

Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.

Qwen3 VL 4B Thinking Versions

Available versions of this model. The score column identifies the version used in the overall ranking.

1 rows
Columns

Show columns

Qwen3 VL 4B ThinkingSep 22, 202523.04B262.1K262.1KNoApache 2.0

What is Qwen3 VL 4B Thinking?

Key information about Qwen3 VL 4B Thinking and its available data.

Qwen3-VL is a large multimodal model that unifies vision, language, and reasoning to achieve human-level perception and cognition across text, images, and video. Built on a 235B-parameter architecture, it integrates early joint training of visual and textual modalities for strong language grounding.

The model supports up to a 1 million-token context window and excels at visual understanding, spatial reasoning, long video comprehension, and tool-based interaction. It can generate code from images, perform precise 2D/3D object grounding, and operate digital interfaces like a visual agent.

5 Pro in perception benchmarks, while the “Thinking” version leads in multimodal reasoning and STEM tasks. With multilingual OCR, creative writing, and fine-grained scene interpretation, Qwen3-VL establishes a new open-source frontier for integrated vision-language intelligence.

Data as of 2026-08-11.

Qwen3 VL 4B Thinking vs nearby models

Open a comparison with the three ranked models immediately above and below this model.

Qwen3 VL 4B ThinkingvsGemma 4 E4BQwen3 VL 4B ThinkingvsDevstral MediumQwen3 VL 4B ThinkingvsGPT-4oQwen3 VL 4B ThinkingvsLlama 3.3 Nemotron Super 49BQwen3 VL 4B ThinkingvsQwen2.5 VL 72BQwen3 VL 4B ThinkingvsMiniCPM SALA

Models similar to Qwen3 VL 4B Thinking

Recommendations prioritize the same model type and family, then the closest LLMBoard score.

#181-0.3
AC

Qwen2.5 VL 72B

Alibaba Cloud / Qwen Team

22.7 LLMBoard

DetailsCompare
#185-1.6
AC

Qwen2.5 72B

Alibaba Cloud / Qwen Team

21.4 LLMBoard

DetailsCompare
#173+1.8
AC

Qwen3 VL 8B

Alibaba Cloud / Qwen Team

24.8 LLMBoard

DetailsCompare
#172+1.8
AC

QwQ 32B

Alibaba Cloud / Qwen Team

24.8 LLMBoard

DetailsCompare
#170+2.5
AC

Qwen2 VL 72B

Alibaba Cloud / Qwen Team

25.5 LLMBoard

DetailsCompare
#194-3.7
AC

Qwen3 VL 4B

Alibaba Cloud / Qwen Team

19.3 LLMBoard

DetailsCompare

FAQ

Common questions about Qwen3 VL 4B Thinking.

When was Qwen3 VL 4B Thinking released?

Qwen3 VL 4B Thinking's default version was released on Sep 22, 2025.

How much does Qwen3 VL 4B Thinking cost?

No official standard PAYG price is currently available for Qwen3 VL 4B Thinking.

Who created Qwen3 VL 4B Thinking?

Qwen3 VL 4B Thinking was created by Alibaba Cloud / Qwen Team.

What is the context window for Qwen3 VL 4B Thinking?

The default version has a 262.1K token context window.

Is Qwen3 VL 4B Thinking open weight?

No. The default version is not marked as having publicly available weights.

How many API providers offer Qwen3 VL 4B Thinking?

No provider offering is currently linked to the default version.

What models should I compare Qwen3 VL 4B Thinking with?

Nearby ranked alternatives include Gemma 4 E4B, Devstral Medium, GPT-4o.