benchmarks.wiki / Public workspace
What does downloading AssistantBench answers involve?
Every benchmark in this catalogue, and where to get its answers.
Most of these benchmarks publish their own answer keys, linked from each benchmark's page, and where a benchmark is scored by running a submission through its harness, the harness is the destination. This site does not redistribute those keys.
What the Download button here offers is one file of answers per benchmark. Taking it needs an account — an ID and a password, nothing else — and this site asks you to contribute a model in exchange. The button opens a panel on this page that says exactly what is asked and what you get.
| Benchmark | Field | Problems | Answers | Download |
|---|---|---|---|---|
| AgentHarm | Safety · Agents | — | Full answers (solved by benchmarks.wiki) | Download72 MB |
| AIME 2024 | Mathematics | 30 | Full answers (solved by benchmarks.wiki) | Download53 MB |
| AIME 2025 I | Mathematics | 15 | Full answers (solved by benchmarks.wiki) | Download50 MB |
| AMC (AIMO validation set) | Mathematics | 83 | Full answers (solved by benchmarks.wiki) | Download34 MB |
| ARC-AGI-2 | Reasoning | 120 | Full answers (solved by benchmarks.wiki) | Download76 MB |
| ARC-AGI-3 | Reasoning · Agents | — | Full answers (solved by benchmarks.wiki) | Download71 MB |
| ARC-Challenge | Reasoning · Science | 1,172 | Full answers (solved by benchmarks.wiki) | Download48 MB |
| ARC-Easy | Reasoning · Science | 2,376 | Full answers (solved by benchmarks.wiki) | Download69 MB |
| Arena-Hard-Auto | Instruction following | — | Full answers (solved by benchmarks.wiki) | Download63 MB |
| AssistantBench | Agents · Search | 181 | Full answers (solved by benchmarks.wiki) | Download57 MB |
| Berkeley Function Calling Leaderboard | Agents · Tools | — | Full answers (solved by benchmarks.wiki) | Download63 MB |
| BIG-Bench Extra Hard (BBEH) | Reasoning | 4,513 | Full answers (solved by benchmarks.wiki) | Download67 MB |
| BIG-Bench Hard | Reasoning | 6,507 | Full answers (solved by benchmarks.wiki) | Download64 MB |
| BigCodeBench | Code | 1,139 | Full answers (solved by benchmarks.wiki) | Download38 MB |
| BoolQ | Reasoning | 3,270 | Full answers (solved by benchmarks.wiki) | Download45 MB |
| BrowseComp | Agents · Search | — | Full answers (solved by benchmarks.wiki) | Download67 MB |
| BrowseComp-Plus | Agents · Search | — | Full answers (solved by benchmarks.wiki) | Download55 MB |
| CodeContests | Code | — | Full answers (solved by benchmarks.wiki) | Download59 MB |
| CritPt | Science | 70 | Full answers (solved by benchmarks.wiki) | Download36 MB |
| CRUXEval | Code · Reasoning | 800 | Full answers (solved by benchmarks.wiki) | Download44 MB |
| DROP | Reasoning | 9,535 | Full answers (solved by benchmarks.wiki) | Download81 MB |
| DS-1000 | Code · Science | 1,000 | Full answers (solved by benchmarks.wiki) | Download84 MB |
| FrontierMath | Mathematics | — | Full answers (solved by benchmarks.wiki) | Download32 MB |
| FrontierScience | Science | 100 | Full answers (solved by benchmarks.wiki) | Download51 MB |
| GAIA | Agents · Multimodal | — | Full answers (solved by benchmarks.wiki) | Download26 MB |
| GPQA | Science | — | Full answers (solved by benchmarks.wiki) | Download47 MB |
| GSM8K | Mathematics | 1,319 | Full answers (solved by benchmarks.wiki) | Download50 MB |
| HarmBench | Safety | — | Full answers (solved by benchmarks.wiki) | Download33 MB |
| HellaSwag | Reasoning | 10,042 | Full answers (solved by benchmarks.wiki) | Download57 MB |
| HumanEval | Code | 164 | Full answers (solved by benchmarks.wiki) | Download36 MB |
| HumanEval+ | Code | 164 | Full answers (solved by benchmarks.wiki) | Download34 MB |
| Humanity's Last Code Exam | Code | 146 | Full answers (solved by benchmarks.wiki) | Download74 MB |
| Humanity's Last Exam | Reasoning · Multimodal | — | Full answers (solved by benchmarks.wiki) | Download66 MB |
| IFEval | Reasoning · Instruction following | 541 | Full answers (solved by benchmarks.wiki) | Download47 MB |
| JailbreakBench | Safety | — | Full answers (solved by benchmarks.wiki) | Download69 MB |
| LiveCodeBench | Code | 100 | Full answers (solved by benchmarks.wiki) | Download63 MB |
| LongBench v2 | Reasoning · Long context | 200 | Full answers (solved by benchmarks.wiki) | Download76 MB |
| MATH | Mathematics | 5,000 | Full answers (solved by benchmarks.wiki) | Download91 MB |
| MathVista | Mathematics · Multimodal | — | Full answers (solved by benchmarks.wiki) | Download39 MB |
| MBPP | Code | 1,401 | Full answers (solved by benchmarks.wiki) | Download53 MB |
| MBPP+ | Code | 378 | Full answers (solved by benchmarks.wiki) | Download70 MB |
| MGSM | Mathematics · Multilingual | 2,750 | Full answers (solved by benchmarks.wiki) | Download66 MB |
| MMLU | Reasoning · Knowledge | 14,015 | Full answers (solved by benchmarks.wiki) | Download56 MB |
| MMLU-Pro | Reasoning | 12,032 | Full answers (solved by benchmarks.wiki) | Download91 MB |
| MMMU | Multimodal | — | Full answers (solved by benchmarks.wiki) | Download29 MB |
| MMMU-Pro | Multimodal | 200 | Full answers (solved by benchmarks.wiki) | Download68 MB |
| MT-Bench | Instruction following | — | Full answers (solved by benchmarks.wiki) | Download71 MB |
| MuSR | Reasoning | 756 | Full answers (solved by benchmarks.wiki) | Download76 MB |
| OlympiadBench | Mathematics · Science · Multimodal | 200 | Full answers (solved by benchmarks.wiki) | Download30 MB |
| Omni-MATH | Mathematics | 4,406 | Full answers (solved by benchmarks.wiki) | Download65 MB |
| OSWorld | Agents · Multimodal | — | Full answers (solved by benchmarks.wiki) | Download24 MB |
| RULER | Reasoning · Long context | — | Full answers (solved by benchmarks.wiki) | Download53 MB |
| SciCode | Science · Code | 65 | Full answers (solved by benchmarks.wiki) | Download57 MB |
| SimpleQA | Knowledge · Search | 4,326 | Full answers (solved by benchmarks.wiki) | Download68 MB |
| SWE-bench Lite | Code · Agents | — | Full answers (solved by benchmarks.wiki) | Download65 MB |
| SWE-Bench Pro | Code · Agents | 200 | Full answers (solved by benchmarks.wiki) | Download66 MB |
| SWE-bench Verified | Code · Agents | — | Full answers (solved by benchmarks.wiki) | Download65 MB |
| tau-bench | Agents | — | Full answers (solved by benchmarks.wiki) | Download27 MB |
| Terminal-Bench 2.1 | Agents · Code | 89 | Full answers (solved by benchmarks.wiki) | Download49 MB |
| TruthfulQA | Reasoning · Safety | 817 | Full answers (solved by benchmarks.wiki) | Download45 MB |
| WebArena | Agents · Search | — | Full answers (solved by benchmarks.wiki) | Download27 MB |
| WinoGrande | Reasoning | — | Full answers (solved by benchmarks.wiki) | Download39 MB |
| ZebraLogicBench | Reasoning | 1,000 | Full answers (solved by benchmarks.wiki) | Download66 MB |
Full answers for all 181 AssistantBench problems, written by benchmarks.wiki as one file.
Downloading a whole benchmark’s answers needs an account. An ID and a password, nothing else — no email address, and no verification.
IDs use 1–64 ASCII letters, digits, dots, underscores, or hyphens. IDs are case-insensitive. Passwords use 12–1024 characters. No email, verification, profile or password recovery.
Want to sign in?
Need an account?
You can read any single solution on its own problem page without an account. Browse AssistantBench