- WildClawBench-MM Multimodal tool use (multimodal tasks only, Claude Code) 71
- UniClawBench Multimodal tool use (OpenClaw) 69.6
- AgenticVBench Multimodal tool use (Claude Code) 36.8
- OmniGAIA Web Search 74
- DailyOmni Audio-Visual Understanding 85.1
- WorldSense Audio-Visual Understanding 68.5
- AVUT Audio-Visual Understanding 86.6
- JoinAVBench Audio-Visual Understanding 75.9
- OmniVideoBench Audio-Visual Reasoning (static; Qwen Code 67.8) 63.4
- Video-MME-v2 Audio-Visual Reasoning (static, audio input; Qwen Code 71.3) 65
- LVOmniBench Long Video Reasoning (static; Qwen Code 73.6) 63.3
- OmniCap-IF Audio-Visual Caption CSR 80.6 ISR 28.2 80.6
- QIVD Audio-Visual Interaction 69.6
- StreamingBench Audio-Visual Interaction 80.8
- FLEURS-ASR Multilingual ASR (WER, 60 languages); FLEURS-S2TT BLEU 31.8 9.3
- SpotSoundBench Audio Grounding 67.2
- MMAU Audio Understanding 81.8
- MMAR Audio Understanding 79.8
- MMSU Audio Understanding 82.1
- LongAudioSpan Long Audio Reasoning Accuracy 82.7 Rubric 71.8 Chain 48.2 82.7
- MuchoMusic-RUL Music Understanding 72.6
- HumMusQA Music Understanding 75.8
- MusTBench Music Understanding 50.6
- Audio MultiChallenge Audio Interaction 71.5
- WildSpeech Audio Interaction 74.3
- VoiceBench Audio Interaction 91.6
- DeepSWE 1.1 (best of Claude Code and mini-SWE-agent) 57.8
- SWE-bench Pro (Claude Code, Qwen-corrected tasks) 63.3
- SWE-bench Multilingual (mini-SWE-agent) 80.5
- NL2Repo-Bench (Claude Code) 48.9
- CoWorkBench Long-horizon office work 75.3
- IFBench 81.5
- GPQA Diamond 91
- HLE (judged by GPT-4o) 36.5
- LiveCodeBench v6 92.6
- ClawEval-MM Pass@3 60.4 Average 61.9 61.9
- AndroidWorld 87.1
- Vision2Web 62.9
- ERQA 71
- LVBench 76.9
- RealWorldQA 87.7
- MathVision Without CI 91.8 With CI 96.2 96.2
- CharXiv (RQ) Without CI 83.5 With CI 91.4 91.4
- OmniCloze Audio-Visual Caption 63.2
- Mean of WenetSpeech (Net) 4.8 / WenetSpeech (Meeting) 4.6 ASR (WER) 4.7
Qwen3.8-Omni-Flash
Qwen3.8-Omni-Flash ranks first on 6 of the 29 benchmarks with five or more models scored.
49
Benchmarks with a score
6
Ranked first, out of five or more models
15
In the top three
Qwen Sep 2026 Qwen3.8-Omni-Flash
Closed weights Mid-size
154.6
#62 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- IFBench 81.5% #5 /80
- NL2Repo-Bench 48.9% #14 /30
- ERQA 71% #7 /36
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| IFBench | 81.5% | #5 of 80 | Lab model card | |
| Humanity's Last Exam | 36.5% | #70 of 154 | Lab model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GPQA Diamond | 91% | #26 of 302 | Lab model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveCodeBench v6 | 92.6% | #2 of 90 | Lab model card | |
| SWE-bench Multilingual | 80.5% | #15 of 73 | Lab model card | |
| SWE-Bench Pro | 63.3% | #20 of 81 | Lab model card | |
| NL2Repo-Bench | 48.9% | #14 of 30 | Lab model card | |
| DeepSWE | 57.8% | #44 of 56 | Lab model card | |
| Vision2Web | 62.9% | #6 of 7 | Lab model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AndroidWorld | 87.1% | #1 of 23 | Lab model card | |
| WildClawBench | 71% | #1 of 5 | Lab model card | |
| OmniGAIA | 74% | #1 of 4 | Lab model card | |
| DuelLab GameBench 2 | 69.7 | #21 of 57 | Official leaderboard | |
| UniClawBench | 69.6% | #1 of 1 | Lab model card |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CoWorkBench | 75.3% | #2 of 7 | Lab model card | |
| YC-Bench | $1,380,061 | #22 of 59 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CharXiv | 91.4% | #4 of 75 | Lab model card | |
| MATH-Vision (MathVision) | 96.2% | #3 of 51 | Lab model card | |
| Daily-Omni | 85.1% | #1 of 14 | Official leaderboard | |
| RealWorldQA | 87.7% | #3 of 38 | Lab model card | |
| VoxMem | 38.5% | #1 of 9 | Official leaderboard | |
| MMAU | 81.8% | #2 of 15 | Lab model card | |
| VoiceBench | 91.6% | #2 of 12 | Lab model card | |
| Audio MC | 71.5% | #1 of 6 | Lab model card | |
| WorldSense (omnimodal) | 68.5% | #1 of 6 | Lab model card | |
| ERQA | 71% | #7 of 36 | Lab model card | |
| LVBench | 76.9% | #7 of 31 | Lab model card | |
| MuChoMusic | 72.6% | #1 of 4 | Lab model card | |
| QIVD | 69.6% | #1 of 4 | Lab model card | |
| VideoMME v2 (with subtitles) | 65% | #2 of 6 | Lab model card | |
| AV-SpeakerBench | 77.2% | #1 of 3 | Lab model card | |
| AVUT | 86.6% | #1 of 3 | Lab model card | |
| Claw-Eval-MM | 61.9% | #3 of 6 | Lab model card | |
| MMSU | 82.1% | #3 of 6 | Lab model card | |
| MMAR | 79.8% | #2 of 4 | Lab model card | |
| JointAVBench | 75.9% | #1 of 2 | Lab model card | |
| OmniVideoBench | 63.4% | #1 of 2 | Lab model card | |
| Omni-Cloze | 63.2% | #2 of 3 | Lab model card | |
| StreamingBench | 80.8% | #2 of 3 | Lab model card | |
| WildSpeech-Bench | 74.3 | #2 of 3 | Lab model card | |
| DECIDE | 30.5 | #6 of 8 | Official leaderboard | |
| AgenticVBench | 36.8% | #1 of 1 | Lab model card | |
| HumMusQA | 75.8% | #1 of 1 | Lab model card | |
| LongAudioSpan | 82.7% | #1 of 1 | Lab model card | |
| LVOmniBench | 63.3% | #1 of 1 | Lab model card | |
| MMOU | 78.8% | #1 of 1 | Lab model card | |
| MusTBench | 50.6% | #1 of 1 | Lab model card | |
| OmniCap-IF | 80.6% | #1 of 1 | Lab model card | |
| SpotSoundBench | 67.2% | #1 of 1 | Lab model card |
Its lab's numbers put it at 154.6, number 62 of 404
The Self-Reported Index fits one score to the 15 capability results Qwen reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 152.1 to 157.1. How we compute it.
- MATH-Vision (MathVision) 96.2
- LiveCodeBench v6 92.6
- VoiceBench 91.6
- CharXiv 91.4
- GPQA Diamond 91
- RealWorldQA 87.7
- AndroidWorld 87.1
- MMAU 81.8
- IFBench 81.5
- SWE-bench Multilingual 80.5
- LVBench 76.9
- SWE-Bench Pro 63.3
- DeepSWE 57.8
- NL2Repo-Bench 48.9
- Humanity's Last Exam 36.5
Its lab's number next to a leaderboard
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| Daily-Omni | 85.1% DailyOmni Audio-Visual Understanding | 85.1% Leaderboard, Omni-Modal Language Models (With Visual and Audio) | Same |
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.
Its system card
Every result in the longer documents the lab published with it. 2 of them aren't in the launch post.
Qwen3.8-Omni: Towards Native Omni-Modal Agents Sep 22, 2026
Capabilities
- DeepSWE 57.8
- SWE-Bench Pro 63.3
- SWE-bench Multilingual 80.5
- NL2Repo-Bench 48.9
- CoWorkBench 75.3
- IFBench 81.5
- GPQA Diamond 91
- Humanity's Last Exam 36.5
- LiveCodeBench v6 92.6
- Claw-Eval-MM 61.9
- AndroidWorld 87.1
- Vision2Web 62.9
- ERQA 71
- LVBench 76.9
- RealWorldQA 87.7
- MATH-Vision (MathVision) 91.8
- CharXiv 83.5
- WenetSpeech
- FLEURS 9.3
- SpotSoundBench 67.2
- MMAU 81.8
- MMAR 79.8
- MMSU 82.1
- LongAudioSpan 82.7
- MuChoMusic 72.6
- HumMusQA 75.8
- MusTBench 50.6
- Audio MC 71.5
- WildSpeech-Bench 74.3
- VoiceBench 91.6
- Daily-Omni 85.1
- WorldSense (omnimodal) 68.5
- AVUT 86.6
- JointAVBench 75.9
- AV-SpeakerBench 77.2
- OmniVideoBench 63.4
- VideoMME v2 (with subtitles) 65
- MMOU 78.8
- LVOmniBench 63.3
- Omni-Cloze 63.2
- OmniCap-IF 80.6
- QIVD 69.6
- StreamingBench 80.8
- WildClawBench 71
- UniClawBench 69.6
- AgenticVBench 36.8
- OmniGAIA 74