- AIME-25 94.3
- HMMT25(Feb) 89.7
- BeyondAIME 77
- AMO-Bench 60
- IMO-AnswerBench (w/ code tools) 76.3
- AetherCode 38.2
- LiveCodeBench(v6) 79.5
- LiveCodeBench pro 1930
- GPQA-Diamond 83.8
- PHYBench 41
- BIOBench (internal) 42.3
- KOR-Bench 76.2
- ARC-AGI-1 67.9
- Inverse IFEval 80.3
- MARS-Bench 70.1
- MultiChallenge 66.7
- Collie-Hard 72.6
- EIFBench 48.6
- MMLU 92.3
- MMLU-pro 84.9
- SuperGPQA 64.8
- LPFQA 49.1
- BrowseComp-en 67.6
- BrowseComp-zh 81.3
- GAIA 87.4
- WideSearch 63.8
- HLE(text-only), agentic search 40.9
- MM-BrowseComp 46.3
- HLE-VL 31.5
- SWE-Bench Verified 72.9
- Multi-SWE-Bench 42
- AInstein-SWE-Bench 36.7
- Terminal Bench 2.0 45.2
- U-Artifacts 49.2
- BFCL-v4 57.2
- τ2-Bench 72
- DeepConsult 41
- DeepResearchBench 43.5
- DiscoX 74.6
- FinSearchComp(T2&T3) 62.8
- Mean of XpertBench Law 55.2 / Fin 62.0 / Edu 47.9 / Research 31.4 / Humanities 60.2 51.3
- WorldTravel multi-modal (best of 5) 47.2 / text 52.1 47.2
- OSWorld 61.9
- Online-Mind2web 85.9
- AndroidWorld 70.7
- MMMU 83.4
- MMMU-Pro 73.2
- MathVista 87.7
- MathVision 81.3
- DynaMath 61.5
- LogicVista 78.3
- ZeroBench (main) 11
- VPCT 61
- VLMsAreBiased 62
- VLMsAreBlind 93
- SimpleVQA 65.4
- HallusionBench 63.9
- MMStar 79.9
- MMBench v1.1 EN (CN 90.6) 91.6
- MUIRBench 78.7
- MMVP 86
- ScreenSpot-Pro 64.3 / 73.1 with crop-box tool 73.1
- CountBench 96.3
- BLINK 74.3
- RefSpatialBench 56.3
- ERQA 58.8
- CV-Bench 88
- AI2D 89.1
- CharXiv (RQ) 71.4
- DUDE 69.4
- Mean of MMLB-NIAH (128k) 72.2 / MMLB-VRAG (128k) 83.2 77.7
- VideoMMMU 82.7
- MMVU 73.1
- VideoHolmes (with subtitles) 65.5
- Minerva 62.4
- VideoSimpleQA 67.8
- TVBench 71.5
- TempCompass 86.9
- TOMATO 60.8
- EgoTempo 67
- MotionBench 70.6
- VideoMME (with subtitles) 87.8
- LongVideoBench 77.4
- LVBench 73
- OVBench 65.1
- OVOBench 72.6
- StreamingBench (proactive) 84.4
- OmniDocBench 1.5 (normalized edit distance, lower is better)
- SFE 51.2
- EMMA 60.9
- MME-CC 43.4
- Point-Bench 76.5
- MMSIBench (circular) 25.8
- VCRBench 59.8
- VideoReasonBench 52.8
- CGBench 62.4
- LiveSports-3K 77.5
- ViSpeak 79
- OmniMMI (proactive) 53
- Education 60.8
- Customer Support Q&A 69
- Information Processing 53.9
- Intention Recognition 63.2
- Information Extraction 43.7
- Complex Workflow 54.6
Seed1.8
Seed1.8 ranks first on 3 of the 75 benchmarks with five or more models scored.
112
Benchmarks with a score
3
Ranked first, out of five or more models
14
In the top three
ByteDance Seed Dec 2025 Seed1.8
Closed weights Flagship
146.0
#123 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- MMMU-Pro 73.2% #48 /90
- BFCL V4 57.2% #22 /30
- SimpleVQA 65.4% #14 /30
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 | |
|---|---|---|---|---|
| MMLU | 92.3% | #1 of 131 | Lab model card | |
| MMLU-Pro | 84.9% | #33 of 167 | Lab model card | |
| MultiChallenge | 66.7% | #9 of 42 | Lab model card | |
| Inverse IFEval | 80.3% | #1 of 3 | Lab model card | |
| SuperGPQA | 64.8% | #14 of 35 | Lab model card | |
| Humanity's Last Exam | 40.9% | #63 of 154 | Lab model card | |
| SimpleVQA | 65.4% | #14 of 30 | Lab model card | |
| MMLongBench | 77.7% | #2 of 4 | Lab model card | |
| HallusionBench | 63.9% | #16 of 31 | Lab model card | |
| COLLIE | 72.6% | #7 of 12 | Lab model card | |
| AMemGym | 41.6% | #7 of 11 | Official leaderboard | |
| CL-bench | 16.1% | #16 of 24 | Official leaderboard | |
| WideSearch | 63.8% | #14 of 19 | Lab model card | |
| Video SimpleQA | 67.8% | #4 of 4 | Lab model card | |
| Complex Workflow (ByteDance Seed in-house) | 54.6% | #2 of 2 | Lab model card | |
| LPFQA | 49.1% | #2 of 2 | Lab model card | |
| MARS-Bench | 70.1% | #2 of 2 | Lab model card | |
| EIFBench | 48.6% | #1 of 1 | Lab model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AIME 2025 | 94.3% | #25 of 143 | Lab model card | |
| Beyond AIME | 77% | #4 of 14 | Lab model card | |
| GPQA Diamond | 83.8% | #91 of 302 | Lab model card | |
| HMMT 2025 | 89.7% | #23 of 54 | Lab model card | |
| AMO-Bench | 60% | #2 of 4 | Lab model card | |
| KOR-Bench | 76.2% | #2 of 4 | Lab model card | |
| ARC-AGI | 67.9% | #14 of 21 | Lab model card | |
| PHYBench | 41% | #4 of 5 | Lab model card | |
| IMO-AnswerBench | 76.3% | #31 of 37 | Lab model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveCodeBench v6 | 79.5% | #33 of 90 | Lab model card | |
| SWE-bench Verified | 72.9% | #60 of 155 | Lab model card | |
| Multi-SWE-bench | 42% | #5 of 10 | Lab model card | |
| LiveCodeBench Pro | 1,930 | #4 of 4 | Lab model card | |
| AetherCode | 38.2% | #3 of 3 | Lab model card | |
| AInstein-SWE-Bench | 36.7% | #2 of 2 | Lab model card | |
| U-Artifacts | 49.2% | #1 of 1 | Lab model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BrowseComp-ZH | 81.3% | #2 of 19 | Lab model card | |
| GAIA | 87.4% | #1 of 5 | Lab model card | |
| OSWorld | 61.9% | #3 of 13 | Lab model card | |
| Online-Mind2Web | 85.9% | #3 of 10 | Lab model card | |
| ScreenSpot-Pro | 73.1% | #11 of 32 | Lab model card | |
| AndroidWorld | 70.7% | #9 of 23 | Lab model card | |
| LisanBench | 1087.3 | #55 of 133 | Official leaderboard | |
| OSWorld-Verified | 61.9% | #15 of 33 | Official leaderboard | |
| GameWorld | 39.8% | #12 of 22 | Official leaderboard | |
| MM-BrowseComp | 46.3% | #3 of 5 | Lab model card | |
| BrowseComp | 67.6% | #49 of 81 | Lab model card | |
| VITA-Bench | 18.8% | #11 of 18 | Official leaderboard | |
| MobileWorld | 36.8% | #12 of 18 | Official leaderboard | |
| tau2-bench | 72% | #55 of 79 | Lab model card | |
| BFCL V4 | 57.2% | #22 of 30 | Lab model card | |
| Terminal-Bench 2.0 | 45.2% | #47 of 63 | Lab model card | |
| Customer Support Q&A (ByteDance Seed in-house) | 69% | #2 of 2 | Lab model card | |
| DeepConsult | 41% | #2 of 2 | Lab model card | |
| DeepResearch Bench | 43.5% | #2 of 2 | Lab model card | |
| WorldTravel | 47.2% | #1 of 1 | Lab model card |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FinSearchComp | 62.8% | #6 of 7 | Lab model card | |
| XpertBench | 51.3% | #3 of 3 | Lab model card | |
| Education (ByteDance Seed in-house) | 60.8% | #2 of 2 | Lab model card | |
| Information Extraction (ByteDance Seed in-house) | 43.7% | #2 of 2 | Lab model card | |
| Information Processing (ByteDance Seed in-house) | 53.9% | #1 of 1 | Lab model card | |
| Intention Recognition (ByteDance Seed in-house) | 63.2% | #1 of 1 | Lab model card |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CVBench | 88% | #1 of 15 | Lab model card | |
| MMMU | 83.4% | #8 of 103 | Lab model card | |
| MathVista | 87.7% | #6 of 70 | Lab model card | |
| Video-MME | 87.8% | #7 of 51 | Lab model card | |
| BLINK | 74.3% | #4 of 29 | Lab model card | |
| LogicVista | 78.3% | #3 of 14 | Lab model card | |
| MMStar | 79.9% | #9 of 36 | Lab model card | |
| MuirBench | 78.7% | #3 of 12 | Lab model card | |
| TempCompass | 86.9% | #2 of 7 | Lab model card | |
| MMBench-EN v1.1 | 91.6% | #7 of 23 | Lab model card | |
| CountBench | 96.3% | #9 of 28 | Lab model card | |
| StreamingBench | 84.4% | #1 of 3 | Lab model card | |
| MMVU | 73.1% | #10 of 26 | Lab model card | |
| MMVP | 86% | #2 of 5 | Lab model card | |
| MATH-Vision (MathVision) | 81.3% | #21 of 51 | Lab model card | |
| AI2D | 89.1% | #22 of 52 | Lab model card | |
| Video-MMMU | 82.7% | #16 of 36 | Lab model card | |
| MotionBench | 70.6% | #4 of 9 | Lab model card | |
| LongVideoBench | 77.4% | #5 of 11 | Lab model card | |
| VLMs Are Blind | 93% | #7 of 14 | Lab model card | |
| LVBench | 73% | #16 of 31 | Lab model card | |
| MMMU-Pro | 73.2% | #48 of 90 | Lab model card | |
| ERQA | 58.8% | #20 of 36 | Lab model card | |
| EgoTempo | 67% | #3 of 5 | Lab model card | |
| OVBench | 65.1% | #3 of 5 | Lab model card | |
| TOMATO | 60.8% | #3 of 5 | Lab model card | |
| RefSpatial-Bench | 56.3% | #10 of 16 | Lab model card | |
| DynaMath | 61.5% | #15 of 23 | Lab model card | |
| BabyVision | 30.2% | #18 of 27 | Official leaderboard | |
| Video-Holmes | 65.5% | #4 of 6 | Lab model card | |
| Point-Bench | 76.5% | #2 of 3 | Lab model card | |
| ViSpeak-Bench | 79% | #2 of 3 | Lab model card | |
| CharXiv | 71.4% | #55 of 75 | Lab model card | |
| DUDE | 69.4% | #3 of 4 | Lab model card | |
| LiveSports-3K | 77.5% | #3 of 4 | Lab model card | |
| OVO-Bench | 72.6% | #4 of 5 | Lab model card | |
| Scientists' First Exam (SFE) | 51.2% | #4 of 5 | Lab model card | |
| MINERVA | 62.4% | #5 of 6 | Lab model card | |
| ZeroBench | 11% | #26 of 31 | Lab model card | |
| TVBench | 71.5% | #6 of 7 | Lab model card | |
| MMSI-Bench | 25.8% | #6 of 6 | Lab model card | |
| VLMsAreBiased | 62% | #5 of 5 | Lab model card | |
| CG-Bench | 62.4% | #3 of 3 | Lab model card | |
| EMMA | 60.9% | #3 of 3 | Lab model card | |
| HLE-VL | 31.5% | #3 of 3 | Lab model card | |
| VideoReasonBench | 52.8% | #3 of 3 | Lab model card | |
| MME-CC | 43.4% | #2 of 2 | Lab model card | |
| VPCT | 61% | #2 of 2 | Lab model card | |
| ZeroVideo | 6.90% | #2 of 2 | Lab model card | |
| OmniMMI | 53% | #1 of 1 | Lab model card | |
| VCR-Bench | 59.8% | #1 of 1 | Lab model card |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BIOBench (ByteDance Seed) | 42.3% | #2 of 2 | Lab model card |
Its lab's numbers put it at 146.0, number 123 of 404
The Self-Reported Index fits one score to the 48 capability results ByteDance Seed 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 144.6 to 147.4. How we compute it.
- CountBench 96.3
- AIME 2025 94.3
- VLMs Are Blind 93
- MMLU 92.3
- MMBench-EN v1.1 91.6
- HMMT 2025 89.7
- AI2D 89.1
- CVBench 88
- Video-MME 87.8
- MathVista 87.7
- MMLU-Pro 84.9
- GPQA Diamond 83.8
- MMMU 83.4
- Video-MMMU 82.7
- BrowseComp-ZH 81.3
- MATH-Vision (MathVision) 81.3
- MMStar 79.9
- LiveCodeBench v6 79.5
- MuirBench 78.7
- LogicVista 78.3
- LongVideoBench 77.4
- Beyond AIME 77
- IMO-AnswerBench 76.3
- BLINK 74.3
- MMMU-Pro 73.2
- ScreenSpot-Pro 73.1
- MMVU 73.1
- LVBench 73
- SWE-bench Verified 72.9
- COLLIE 72.6
- tau2-bench 72
- CharXiv 71.4
- AndroidWorld 70.7
- MotionBench 70.6
- ARC-AGI 67.9
- BrowseComp 67.6
- MultiChallenge 66.7
- SimpleVQA 65.4
- SuperGPQA 64.8
- HallusionBench 63.9
- WideSearch 63.8
- OSWorld 61.9
- DynaMath 61.5
- BFCL V4 57.2
- Terminal-Bench 2.0 45.2
- Multi-SWE-bench 42
- Humanity's Last Exam 40.9
- ZeroBench 11
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 |
|---|---|---|---|
| MATH-Vision (MathVision) | 81.3% MathVision | 81.3% Leaderboard | 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. 3 of them aren't in the launch post.
Capabilities
- AIME 2025 94.3
- HMMT 2025 89.7
- Beyond AIME 77
- AMO-Bench 60
- IMO-AnswerBench 76.3
- AetherCode 38.2
- LiveCodeBench v6 79.5
- LiveCodeBench Pro 1930
- GPQA Diamond 83.8
- PHYBench 41
- BIOBench (ByteDance Seed) 42.3
- KOR-Bench 76.2
- ARC-AGI 67.9
- Inverse IFEval 80.3
- MARS-Bench 70.1
- MultiChallenge 66.7
- COLLIE 72.6
- EIFBench 48.6
- MMLU 92.3
- MMLU-Pro 84.9
- SuperGPQA 64.8
- LPFQA 49.1
- Education (ByteDance Seed in-house) 60.8
- Customer Support Q&A (ByteDance Seed in-house) 69
- Information Processing (ByteDance Seed in-house) 53.9
- Intention Recognition (ByteDance Seed in-house) 63.2
- Information Extraction (ByteDance Seed in-house) 43.7
- Complex Workflow (ByteDance Seed in-house) 54.6
- MMMU 83.4
- MMMU-Pro 73.2
- MathVista 87.7
- MATH-Vision (MathVision) 81.3
- DynaMath 61.5
- LogicVista 78.3
- EMMA 60.9
- Scientists' First Exam (SFE) 51.2
- ZeroBench 11
- VPCT 61
- VLMsAreBiased 62
- VLMs Are Blind 93
- SimpleVQA 65.4
- HallusionBench 63.9
- MMStar 79.9
- MMBench-EN v1.1 91.6
- MME-CC 43.4
- MuirBench 78.7
- MMVP 86
- ScreenSpot-Pro 64.3
- CountBench 96.3
- Point-Bench 76.5
- BLINK 74.3
- MMSI-Bench 25.8
- RefSpatial-Bench 56.3
- ERQA 58.8
- CVBench 88
- AI2D 89.1
- OmniDocBench v1.5
- CharXiv 71.4
- DUDE 69.4
- MMLongBench
- Video-MMMU 82.7
- MMVU 73.1
- VCR-Bench 59.8
- VideoReasonBench 52.8
- Video-Holmes 65.5
- MINERVA 62.4
- Video SimpleQA 67.8
- TVBench 71.5
- TempCompass 86.9
- TOMATO 60.8
- EgoTempo 67
- MotionBench 70.6
- Video-MME 87.8
- CG-Bench 62.4
- LongVideoBench 77.4
- LVBench 73
- OVBench 65.1
- LiveSports-3K 77.5
- OVO-Bench 72.6
- ViSpeak-Bench 79
- StreamingBench 84.4
- OmniMMI 53
- ZeroVideo 6.9
- BrowseComp 67.6
- BrowseComp-ZH 81.3
- GAIA 87.4
- WideSearch 63.8
- Humanity's Last Exam 21.4
- MM-BrowseComp 46.3
- HLE-VL 17.3
- SWE-bench Verified 72.9
- Multi-SWE-bench 42
- AInstein-SWE-Bench 36.7
- Terminal-Bench 2.0 45.2
- U-Artifacts 49.2
- BFCL V4 57.2
- tau2-bench 72
- DeepConsult 41
- DeepResearch Bench 43.5
- DiscoX 74.6
- FinSearchComp 62.8
- XpertBench
- WorldTravel
- OSWorld 61.9
- Online-Mind2Web 85.9
- AndroidWorld 70.7