AIME 2025
Harbor adapterAIME 2025: 30道AIME级数学题
Verified86%Math
- Samples
- 30
- Released
- 2025
- Scores
- 12
IFEval
Instruction Following Eval
IFEval: 500+可编程验证指令
How to run
ltb · ifeval
$ npm i -g lettokenburn-cli$ ltb bench pull ifeval
External harness. pip install lm-eval && lm_eval --model <hf|vllm|openai-chat-completions> --model_args <model args> --tasks ifeval (lm-evaluation-harness task family 'ifeval') https://github.com/EleutherAI/lm-evaluation-harness
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
10 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| Claude Opus 4.6 | 94.0 | N/A | Self-reported | low | Jun 1, 2026 | artificialanalysis.ai |
| Kimi K2.5 | 94.0 | N/A | Self-reported | low | Jun 1, 2026 | artificialanalysis.ai |
| Qwen 3.5 | 92.6 | N/A | Self-reported | low | Jun 1, 2026 | artificialanalysis.ai |
| GLM-5 | 88.0 | N/A | Self-reported | low | Jun 1, 2026 | artificialanalysis.ai |
| MiniMax M2.5 | 87.5 | N/A | Self-reported | low | Jun 1, 2026 | artificialanalysis.ai |
| GPT-4 | 85.4 | N/A | Independent | medium | Nov 14, 2023 | arxiv.org |
| Gemini 3.1 Pro | 85.0 | N/A | Self-reported | low | Jun 1, 2026 | artificialanalysis.ai |
| DeepSeek R1 | 83.3 | N/A | Self-reported | low | Jun 1, 2026 | artificialanalysis.ai |
| Gemini 3 Pro | 82.0 | N/A | Self-reported | low | Jun 1, 2026 | artificialanalysis.ai |
| PaLM 2-S | 59.1 | N/A | Independent | medium | Nov 14, 2023 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsLoading community runs…
About
IFEval (Instruction Following Eval) — 500+可编程验证指令. Scoring: 严格/宽松准确率. Difficulty: 中（趋于饱和）. Imported from benchmark_research.md (2026-06).