- MMLU-Pro 87
- HLE-text (search agent, Table 11) 54.2 / HLE (no tool, text only) 32.4 54.2
- HLE-Verified (search agent) 73.6
- SimpleQA Verified 36
- HealthBench 57.7
- HealthBench - Hard 29.1
- SuperGPQA 68.7
- LPFQA 52.6
- Encyclo-K 65.7
- AIME 2026 94.2
- AIME 2025 98.3
- HMMT Feb 2025 97.3
- HMMT Nov 2025 93.3
- MathArenaApex 20.3 (shortlist 82.1) 20.3
- BeyondAIME 86.5
- IMOAnswerBench (no tool) 89.3
- Codeforces (no tool) 3020
- AetherCode 60.6
- LiveCodeBench (v6) 87.8
- GPQA Diamond 88.9
- Superchem (text-only) 51.6
- BABE 50
- Phybench 74
- FrontierSci-olympiad 74.0 (FrontierSci-research 25.0) 74
- ARC-AGI-1 (text) 85.4 / ArcAGI1-Image 88.8 85.4
- ARC-AGI-2 (text) 37.5 / ArcAGI2-Image 43.3 37.5
- KORBench 77.5
- ProcBench 96.6
- MRCR v2 (8-needle) 54
- Mean of Graphwalks Bfs (<128k) 68.9 / Graphwalks Parents (<128k) 97.6 83.3
- LongBench v2 (128k) 63.8
- Frames 84.5
- DeR2 Bench 58.2
- CL-Bench 20.8
- Global PIQA 92.3
- MMMLU 88.1
- Disco-X 82
- MultiChallenge 68.3
- COLLIE 93.9
- MARS-Bench 85.6
- Inverse IFEval 78.9
- Mean of LongFact-Objects 92.9 / LongFact-Concepts 92.8 92.8
- FactScore 71.2
- Terminal Bench 2.0 (Terminus 2, 3 tasks excluded) 55.8
- SWE-Lancer 49.4
- SWE Bench Verified 76.5
- Multi-SWE-Bench 45.2
- SWE-Bench Pro 46.9
- SWE Multilingual 71.7
- Scicode 48.5
- SWE-Evo 8.5
- Aider Polyglot 80
- ArtifactsBench 66.6
- CodeSimpleQA 58
- SpreadsheetBench Verified 79.1
- BrowseComp 77.3
- BrowseComp-zh 82.4
- WideSearch 74.7
- FinSearchComp 70.2
- DeepSearchQA 77.4
- Seal-0 49.5
- Mean of τ2-Bench (retail) 90.4 / τ2-Bench (telecom) 94.2 92.3
- MCP-Mark 54.7
- BFCL-v4 73.4
- VitaBench 47
- DeepConsult 61.1
- DeepResearchBench 53.3
- ResearchRubrics 50.7
- MM-BrowseComp 48.8
- HLE-VL 39.2
- MathVista 89.8
- MathVision 88.8
- LogicVista 81.9
- VisuLogic 47.4
- VLMsAreBiased 77.4
- VLMsAreBlind 98.6
- BabyVision 60.6
- ChartQAPro 71.2
- OmniDocBench 1.5 (normalized edit distance, lower is better) 0.099
- DUDE 72.4
- MMLongBench 74.8
- MMLongBench-Doc 61.4
- TVBench 75
- TempCompass 89.6
- MotionBench 75.2
- EgoTempo 71.8
- VideoMME (with subtitles) 89.5
- LVBench (without VideoCut, 80.0 with) 76.4
- MathKangaroo 90.5
- MathCanvas 61.9
- Minedojo-Verified 49
- Putnam-200 Pass@8, agentic Lean setup (Table 5) 35.5
- IMO 2025, 35/42, gold (Table 2) 35
- CMO 2025, 114/126, gold (Table 2) 114
Seed2.0 Pro
Seed2.0 Pro ranks first on 8 of the 117 benchmarks it has scores on.
178
Benchmarks with a score
8
Ranked first, out of five or more models
40
In the top three
ByteDance Seed Feb 2026 Seed2.0 Pro
Closed weights Flagship
No Epoch Capabilities Index score yet
- MMMU-Pro 78.2% #21 /72
- SciCode 48.5% #13 /47
- NL2Repo-Bench 27.9% #27 /28
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-Pro | 87% | #11 of 151 | Model card | |
| LMArena (Elo) | 1,456 | #53 of 333 | Official leaderboard | |
| Humanity's Last Exam | 54.2% | #26 of 147 | Model card | |
| LongBench v2 | 63.8% | #5 of 27 | Model card | |
| MultiChallenge | 68.3% | #7 of 36 | Model card | |
| HallusionBench | 68% | #5 of 23 | Model card | |
| HLE-Verified | 73.6% | #1 of 4 | Model card | |
| SimpleVQA | 71.4% | #7 of 23 | Model card | |
| SuperGPQA | 68.7% | #8 of 26 | Model card | |
| MMMLU | 88.1% | #20 of 62 | Model card | |
| GraphWalks | 83.3% | #7 of 21 | Model card | |
| FRAMES | 84.5% | #3 of 8 | Model card | |
| Global PIQA | 92.3% | #3 of 8 | Model card | |
| WideSearch | 74.7% | #7 of 17 | Model card | |
| COLLIE | 93.9% | #5 of 12 | Model card | |
| CL-bench Life | 15.5% | #7 of 16 | Official leaderboard | |
| CL-bench | 20.8% | #3 of 6 | Model card | |
| Video SimpleQA | 71.9% | #2 of 4 | Model card | |
| Complex Workflow (ByteDance Seed in-house) | 64.7% | #1 of 2 | Model card | |
| DiscoX | 82% | #1 of 2 | Model card | |
| LPFQA | 52.6% | #1 of 2 | Model card | |
| MARS-Bench | 85.6% | #1 of 2 | Model card | |
| SimpleQA Verified | 36% | #12 of 22 | Model card | |
| Seal-0 | 49.5% | #3 of 5 | Model card | |
| Inverse IFEval | 78.9% | #2 of 3 | Model card | |
| MMLongBench | 74.8% | #3 of 4 | Model card | |
| MRCR v2 | 54% | #21 of 27 | Model card | |
| Extended NYT Connections | 28.4% | #76 of 97 | Official leaderboard | |
| LongFact | 92.8% | #4 of 5 | Model card | |
| FActScore | 71.2% | #4 of 4 | Model card | |
| LongFact-Concepts | 92.8% | #4 of 4 | Model card | |
| LongFact-Objects | 92.9% | #4 of 4 | Model card | |
| Chinese Complex Instruction Following (ByteDance Seed in-house) | 75.3% | #1 of 1 | Model card | |
| CodeSimpleQA | 58% | #1 of 1 | Model card | |
| Encyclo-K | 65.7% | #1 of 1 | Model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AIME 2025 | 98.3% | #14 of 143 | Model card | |
| Beyond AIME | 86.5% | #2 of 14 | Model card | |
| GPQA Diamond | 88.9% | #41 of 284 | Model card | |
| IMO-AnswerBench | 89.3% | #6 of 37 | Model card | |
| HMMT 2025 | 97.3% | #8 of 49 | Model card | |
| KOR-Bench | 77.5% | #1 of 4 | Model card | |
| IMO 2025 | 83.3% | #2 of 6 | Model card | |
| CMO 2025 | 114.0 | #1 of 3 | Model card | |
| PHYBench | 74% | #2 of 5 | Model card | |
| FrontierScience | 74% | #3 of 7 | Model card | |
| HMMT Nov 2025 | 93.3% | #10 of 23 | Model card | |
| AIME 2026 | 94.2% | #16 of 34 | Model card | |
| ARC-AGI | 85.4% | #9 of 19 | Model card | |
| Echo General AI Prediction Leaderboard | 996 | #11 of 17 | Official leaderboard | |
| PhyX | 72.1% | #2 of 3 | Model card | |
| ARC-AGI-2 | 37.5% | #17 of 22 | Model card | |
| xbench-ScienceQA | 69.2% | #23 of 29 | Official leaderboard | |
| PutnamBench | 35.5% | #6 of 7 | Model card | |
| MathArena (Apex, IMO, USAMO) | 20.3% | #13 of 13 | Model card | |
| HiPhO | 74.1% | #4 of 4 | Model card | |
| FrontierScience-Research | 25% | #3 of 3 | Model card | |
| FrontierScience-Olympiad | 74% | #2 of 2 | Model card | |
| ProcBench | 96.6% | #1 of 1 | Model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Codeforces rating | 3,020 | #5 of 41 | Model card | |
| Aider Polyglot | 80% | #4 of 29 | Model card | |
| LiveCodeBench v6 | 87.8% | #11 of 76 | Model card | |
| SWE-bench Verified | 76.5% | #36 of 150 | Model card | |
| SciCode | 48.5% | #13 of 47 | Model card | |
| Multi-SWE-bench | 45.2% | #4 of 10 | Model card | |
| ArtifactsBench | 66.6% | #2 of 5 | Model card | |
| AInstein-SWE-Bench | 47.7% | #1 of 2 | Model card | |
| SWE-bench Multilingual | 71.7% | #39 of 70 | Model card | |
| OJBench | 54.9% | #5 of 8 | Official leaderboard | |
| AetherCode | 60.6% | #2 of 3 | Model card | |
| SWE-Lancer | 49.4% | #8 of 10 | Model card | |
| SWE-Bench Pro | 46.9% | #70 of 77 | Model card | |
| NL2Repo-Bench | 27.9% | #27 of 28 | Model card | |
| EvoCode-Bench | 1.90% | #13 of 13 | Official leaderboard | |
| SWE-EVO | 8.50% | #1 of 1 | Model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BrowseComp-ZH | 82.4% | #1 of 17 | Model card | |
| BFCL V4 | 73.4% | #2 of 24 | Model card | |
| VitaBench 2.0 | 47.4% | #2 of 20 | Official leaderboard | |
| tau2-bench | 92.3% | #11 of 69 | Model card | |
| MobileWorld | 63.2% | #3 of 18 | Official leaderboard | |
| VITA-Bench | 47% | #3 of 12 | Model card | |
| LisanBench | 1753.0 | #47 of 133 | Official leaderboard | |
| MM-BrowseComp | 48.8% | #2 of 5 | Model card | |
| BrowseComp | 77.3% | #35 of 77 | Model card | |
| Terminal-Bench 2.0 | 55.8% | #30 of 61 | Model card | |
| MobileGym | 52% | #2 of 4 | Official leaderboard | |
| Customer Support Q&A (ByteDance Seed in-house) | 72.4% | #1 of 2 | Model card | |
| DeepConsult | 61.1% | #1 of 2 | Model card | |
| DeepResearch Bench | 53.3% | #1 of 2 | Model card | |
| MCPMark | 54.7% | #6 of 10 | Model card | |
| DeepSearchQA | 77.4% | #15 of 22 | Model card | |
| PACT | 1,491 | #21 of 26 | Official leaderboard | |
| ResearchRubrics | 50.7% | #5 of 5 | Model card | |
| DeR2 Bench | 58.2% | #1 of 1 | Model card | |
| MineDojo | 49% | #1 of 1 | Model card |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FinSearchComp | 70.2% | #2 of 7 | Model card | |
| XpertBench | 64.5% | #1 of 3 | Model card | |
| Education (ByteDance Seed in-house) | 62.8% | #1 of 2 | Model card | |
| Information Extraction (ByteDance Seed in-house) | 52% | #1 of 2 | Model card | |
| SpreadsheetBench Verified | 79.1% | #1 of 2 | Model card | |
| HealthBench | 57.7% | #19 of 30 | Model card | |
| FORTE | 57.7% | #18 of 21 | Official leaderboard | |
| HealthBench Hard | 29.1% | #19 of 21 | Model card | |
| Compositional Tasks (ByteDance Seed in-house) | 59.1% | #1 of 1 | Model card | |
| GDPVal-Diamond | 21.3% | #1 of 1 | Model card | |
| Text Classification (ByteDance Seed in-house) | 69% | #1 of 1 | Model card |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MMMU | 85.4% | #3 of 97 | Model card | |
| MathVista | 89.8% | #4 of 64 | Model card | |
| VideoMME v2 (with subtitles) | 43.3% | #2 of 31 | Official leaderboard | |
| Video-MME | 89.5% | #3 of 42 | Model card | |
| BLINK | 79.5% | #2 of 28 | Model card | |
| MuirBench | 81.8% | #1 of 12 | Model card | |
| MotionBench | 75.2% | #1 of 9 | Model card | |
| Vibe-Eval | 81.4% | #1 of 9 | Model card | |
| VLMs Are Blind | 98.6% | #1 of 8 | Model card | |
| MMStar | 83% | #4 of 30 | Model card | |
| TempCompass | 89.6% | #1 of 7 | Model card | |
| Video-MMMU | 86.9% | #5 of 30 | Model card | |
| LogicVista | 81.9% | #2 of 12 | Model card | |
| All-Angles Bench | 72.1% | #1 of 6 | Model card | |
| LongVideoBench | 80.3% | #2 of 11 | Model card | |
| RealWorldQA | 86% | #6 of 30 | Model card | |
| EgoTempo | 71.8% | #1 of 5 | Model card | |
| MATH-Vision (MathVision) | 88.8% | #9 of 42 | Model card | |
| MMVU | 78.2% | #5 of 20 | Model card | |
| ChartQA Pro | 71.2% | #2 of 8 | Model card | |
| RefSpatial-Bench | 72.6% | #2 of 8 | Model card | |
| CharXiv (DQ) | 93.5% | #2 of 7 | Model card | |
| TreeBench | 64.7% | #2 of 7 | Model card | |
| VisuLogic | 47.4% | #2 of 7 | Model card | |
| MMMU-Pro | 78.2% | #21 of 72 | Model card | |
| Vision Arena | 1,256 | #41 of 136 | Official leaderboard | |
| ERQA | 68.5% | #8 of 26 | Model card | |
| MMLongBench-Doc | 61.4% | #4 of 13 | Model card | |
| MTVQA-Test | 51.1% | #2 of 6 | Model card | |
| Video-Holmes | 67.4% | #2 of 6 | Model card | |
| ContPhy | 67.4% | #1 of 3 | Model card | |
| Point-Bench | 81.4% | #1 of 3 | Model card | |
| VideoReasonBench | 77.8% | #1 of 3 | Model card | |
| CountBench | 95.5% | #7 of 20 | Model card | |
| LVBench | 76.4% | #9 of 25 | Model card | |
| OVBench | 69.2% | #2 of 5 | Model card | |
| OVO-Bench | 77% | #2 of 5 | Model card | |
| Scientists' First Exam (SFE) | 55.6% | #2 of 5 | Model card | |
| BabyVision | 60.6% | #11 of 27 | Official leaderboard | |
| MDPBench | 74.2 | #4 of 9 | Official leaderboard | |
| CharXiv | 80.5% | #31 of 67 | Model card | |
| OCRBench v2 | 62.5% | #10 of 20 | Model card | |
| DUDE | 72.4% | #2 of 4 | Model card | |
| LiveSports-3K | 78% | #2 of 4 | Model card | |
| XLRS-Bench | 54.6% | #2 of 4 | Model card | |
| MME-CC | 57% | #1 of 2 | Model card | |
| MORSE-500 | 37.4% | #1 of 2 | Model card | |
| ODV-Bench | 72.5% | #1 of 2 | Model card | |
| VPCT | 76% | #1 of 2 | Model card | |
| ZeroVideo | 14.5% | #1 of 2 | Model card | |
| DynaMath | 68.9% | #9 of 17 | Model card | |
| MMSI-Bench | 32.5% | #3 of 5 | Model card | |
| MINERVA | 66.5% | #4 of 6 | Model card | |
| CG-Bench | 65% | #2 of 3 | Model card | |
| EMMA | 72% | #2 of 3 | Model card | |
| HLE-VL | 39.2% | #2 of 3 | Model card | |
| TVBench | 75% | #5 of 7 | Model card | |
| ZeroBench | 12% | #19 of 25 | Model card | |
| WorldVQA | 49.9% | #7 of 9 | Model card | |
| TOMATO | 59.9% | #4 of 5 | Model card | |
| VLMsAreBiased | 77.4% | #4 of 5 | Model card | |
| OmniDocBench v1.5 | 0.10 | #13 of 13 | Model card | |
| VisFactor | 36.8% | #4 of 4 | Model card | |
| CrossVid | 60.3% | #3 of 3 | Model card | |
| ViSpeak-Bench | 78.5% | #3 of 3 | Model card | |
| VideoEval-Pro | 48% | #2 of 2 | Model card | |
| LongDocURL | 74.7% | #1 of 1 | Model card | |
| Math Kangaroo 2025 (MathArena) | 90.5% | #1 of 1 | Model card | |
| MathCanvas-Bench | 61.9% | #1 of 1 | Model card | |
| ViVerBench | 75.9% | #1 of 1 | Model card |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BIOBench (ByteDance Seed) | 53.5% | #1 of 2 | Model card | |
| SUPERChem | 51.6% | #5 of 5 | Model card | |
| BABE | 50% | #2 of 2 | Model card |
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. 69 of them aren't in the launch post.
Capabilities
- IMO 2025 83.3
- CMO 2025 114
- MMLU-Pro 87
- Humanity's Last Exam 32.4
- SuperGPQA 68.7
- LPFQA 52.6
- Encyclo-K 65.7
- AIME 2026 94.2
- AIME 2025 98.3
- HMMT 2025 97.3
- HMMT Nov 2025 93.3
- MathArena (Apex, IMO, USAMO) 20.3
- Beyond AIME 86.5
- IMO-AnswerBench 89.3
- Codeforces rating 3020
- AetherCode 60.6
- LiveCodeBench v6 87.8
- GPQA Diamond 88.9
- SUPERChem 51.6
- BABE 50
- PHYBench 74
- FrontierScience-Research 25
- FrontierScience-Olympiad 74
- ARC-AGI 85.4
- ARC-AGI-2 37.5
- KOR-Bench 77.5
- ProcBench 96.6
- MRCR v2 54
- GraphWalks
- LongBench v2 63.8
- FRAMES 84.5
- DeR2 Bench 58.2
- CL-bench 20.8
- Global PIQA 92.3
- MMMLU 88.1
- DiscoX 82
- MultiChallenge 68.3
- COLLIE 93.9
- MARS-Bench 85.6
- Inverse IFEval 78.9
- PutnamBench 35.5
- Chinese Complex Instruction Following (ByteDance Seed in-house) 75.3
- MathVista 89.8
- MATH-Vision (MathVision) 88.8
- DynaMath 68.9
- Math Kangaroo 2025 (MathArena) 90.5
- MathCanvas-Bench 61.9
- MMMU 85.4
- MMMU-Pro 78.2
- EMMA 72
- Scientists' First Exam (SFE) 55.6
- HiPhO 74.1
- XLRS-Bench 54.6
- PhyX 72.1
- LogicVista 81.9
- VPCT 76
- ZeroBench 12
- VisuLogic 47.4
- VLMsAreBiased 77.4
- VLMs Are Blind 98.6
- VisFactor 36.8
- RealWorldQA 86
- BabyVision 60.6
- SimpleVQA 71.4
- HallusionBench 68
- MME-CC 57
- MMStar 83
- MuirBench 81.8
- MTVQA-Test 51.1
- WorldVQA 49.9
- Vibe-Eval 81.4
- ViVerBench 75.9
- CountBench 95.5
- Point-Bench 81.4
- BLINK 79.5
- MMSI-Bench 32.5
- TreeBench 64.7
- RefSpatial-Bench 72.6
- All-Angles Bench 72.1
- ERQA 68.5
- ChartQA Pro 71.2
- OCRBench v2 62.5
- OmniDocBench v1.5 0.099
- CharXiv (DQ) 93.5
- CharXiv 80.5
- DUDE 72.4
- MMLongBench 74.8
- LongDocURL 74.7
- MMLongBench-Doc 61.4
- Video-MMMU 86.9
- MMVU 78.2
- Video SimpleQA 71.9
- VideoReasonBench 77.8
- MORSE-500 37.4
- Video-Holmes 67.4
- MINERVA 66.5
- TVBench 75
- ContPhy 67.4
- TempCompass 89.6
- EgoTempo 71.8
- MotionBench 75.2
- TOMATO 59.9
- Video-MME 89.5
- CG-Bench 65
- LongVideoBench 80.3
- VideoEval-Pro 48
- LVBench 76.4
- CrossVid 60.3
- OVBench 69.2
- LiveSports-3K 78
- OVO-Bench 77
- ODV-Bench 72.5
- ViSpeak-Bench 78.5
- ZeroVideo 14.5
- Terminal-Bench 2.0 55.8
- SWE-Lancer 49.4
- SWE-bench Verified 76.5
- Multi-SWE-bench 45.2
- SWE-Bench Pro 46.9
- SWE-bench Multilingual 71.7
- SciCode 48.5
- SWE-EVO 8.5
- Aider Polyglot 80
- ArtifactsBench 66.6
- CodeSimpleQA 58
- SpreadsheetBench Verified 79.1
- BrowseComp 77.3
- BrowseComp-ZH 82.4
- HLE-Verified 73.6
- WideSearch 74.7
- FinSearchComp 70.2
- DeepSearchQA 77.4
- Seal-0 49.5
- tau2-bench
- MCPMark 54.7
- BFCL V4 73.4
- VITA-Bench 47
- DeepConsult 61.1
- DeepResearch Bench 53.3
- ResearchRubrics 50.7
- MineDojo 49
- MM-BrowseComp 48.8
- HLE-VL 39.2
- BIOBench (ByteDance Seed) 53.5
- AInstein-SWE-Bench 47.7
- NL2Repo-Bench 27.9
- Complex Workflow (ByteDance Seed in-house) 64.7
- Customer Support Q&A (ByteDance Seed in-house) 72.4
- GDPVal-Diamond 21.3
- XpertBench 64.5
- Education (ByteDance Seed in-house) 62.8
- Compositional Tasks (ByteDance Seed in-house) 59.1
- Text Classification (ByteDance Seed in-house) 69
- Information Extraction (ByteDance Seed in-house) 52
- WorldTravel
Health
- HealthBench 57.7
- HealthBench Hard 29.1
Hallucination
- SimpleQA Verified 36
- LongFact-Objects 92.9
- LongFact-Concepts 92.8
- FActScore 71.2