| GPT-2 XL (1.5B) | 40.0% | Self-reported | Textbooks Are All You Need II: phi-1.5 technical report via Epoch AI | — |
| StableLM-Alpha 7B | 40.7% | Self-reported | XGen-7B Technical Report via Epoch AI | — |
| OPT-1.3B | 41.5% | Self-reported | Textbooks Are All You Need II: phi-1.5 technical report via Epoch AI | — |
| text-ada-001 (InstructGPT) | 42.9% | Third party | Stanford HELM via Epoch AI | — |
| GPT-3 ada | 43.5% | Third party | Stanford HELM via Epoch AI | — |
| Phi-1.5 | 47.6% | Self-reported | Textbooks Are All You Need II: phi-1.5 technical report via Epoch AI | — |
| Phi-2 | 53.6% | Self-reported | Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI | — |
| Falcon-7B Instruct | 54.2% | Self-reported | Textbooks Are All You Need II: phi-1.5 technical report via Epoch AI | — |
| GPT-3 babbage | 55.5% | Third party | Stanford HELM via Epoch AI | — |
| text-babbage-001 (InstructGPT) | 56.1% | Third party | Stanford HELM via Epoch AI | — |
| ChatGLM2-6B | 57.0% | Self-reported | Qwen Technical Report via Epoch AI | — |
| Llama 2 7B | 57.1% | Self-reported | Textbooks Are All You Need II: phi-1.5 technical report via Epoch AI | — |
| MPT-7B (Instruct, Chat, StoryWriter) | 57.1% | Self-reported | Textbooks Are All You Need II: phi-1.5 technical report via Epoch AI | — |
| text-curie-001 (InstructGPT) | 67.6% | Third party | Stanford HELM via Epoch AI | — |
| GPT-3 curie | 68.2% | Third party | Stanford HELM via Epoch AI | — |
| OPT-13B | 69.9% | Self-reported | XGen-7B Technical Report via Epoch AI | — |
| Dolly 2.0 (3B, 7B, 12B) | 70.8% | Self-reported | XGen-7B Technical Report via Epoch AI | — |
| Gemma 2B | 71.4% | Self-reported | Gemma: Open Models Based on Gemini Research and Technology via Epoch AI | — |
| Falcon-7B Instruct | 74.1% | Self-reported | Qwen Technical Report via Epoch AI | — |
| OPT-66B | 74.5% | Third party | Stanford HELM via Epoch AI | — |
| MPT-7B (Instruct, Chat, StoryWriter) | 76.1% | Self-reported | XGen-7B Technical Report via Epoch AI | — |
| Falcon-7B Instruct | 76.3% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| Falcon-7B Instruct | 76.3% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| Falcon-7B Instruct | 76.4% | Self-reported | XGen-7B Technical Report via Epoch AI | — |
| MPT-7B (Instruct, Chat, StoryWriter) | 76.4% | Self-reported | Qwen Technical Report via Epoch AI | — |
| Yi-9B | 76.4% | Self-reported | Yi: Open Foundation Models by 01.AI via Epoch AI | — |
| Llama 2 7B | 77.2% | Third party | Gemma: Open Models Based on Gemini Research and Technology via Epoch AI | — |
| GPT-3 davinci | 77.5% | Third party | Stanford HELM via Epoch AI | — |
| Falcon-7B Instruct | 78.1% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| Falcon-7B Instruct | 78.1% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| text-davinci-001 (InstructGPT) | 78.1% | Third party | Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model via Epoch AI | — |
| text-davinci-001 (InstructGPT) | 78.1% | Third party | GLaM: Efficient Scaling of Language Models with Mixture-of-Experts via Epoch AI | — |
| text-davinci-001 (InstructGPT) | 78.1% | Third party | Language Models are Few-Shot Learners via Epoch AI | — |
| text-davinci-001 (InstructGPT) | 78.9% | Third party | Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model via Epoch AI | — |
| text-davinci-001 (InstructGPT) | 78.9% | Third party | GLaM: Efficient Scaling of Language Models with Mixture-of-Experts via Epoch AI | — |
| text-davinci-001 (InstructGPT) | 78.9% | Third party | Language Models are Few-Shot Learners via Epoch AI | — |
| text-davinci-001 (InstructGPT) | 79.3% | Third party | Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model via Epoch AI | — |
| text-davinci-001 (InstructGPT) | 79.3% | Third party | GLaM: Efficient Scaling of Language Models with Mixture-of-Experts via Epoch AI | — |
| text-davinci-001 (InstructGPT) | 79.3% | Third party | Language Models are Few-Shot Learners via Epoch AI | — |
| Llama 2 13B | 80.7% | Third party | Gemma: Open Models Based on Gemini Research and Technology via Epoch AI | — |
| Mistral 7B v0.1 | 81.0% | Third party | Mixtral of Experts via Epoch AI | — |
| Gemma 7B | 81.2% | Self-reported | Gemma: Open Models Based on Gemini Research and Technology via Epoch AI | — |
| text-davinci-002 | 81.5% | Third party | Stanford HELM via Epoch AI | — |
| Falcon 2 11B | 82.1% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| Falcon 2 11B | 82.1% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| Gemma 7B | 82.2% | Self-reported | Gemma: Open Models Based on Gemini Research and Technology via Epoch AI | — |
| text-davinci-003 | 82.2% | Third party | Stanford HELM via Epoch AI | — |
| Phi-3-medium | 82.4% | Self-reported | Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI | — |
| Falcon-40B Instruct | 82.8% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| Falcon-40B Instruct | 82.8% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| Falcon 2 11B | 82.9% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| Falcon 2 11B | 82.9% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| Stable Beluga 2 (Llama 2 70B fine-tune) | 84.1% | Self-reported | Qwen Technical Report via Epoch AI | — |
| Falcon-40B Instruct | 85.3% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| Falcon-40B Instruct | 85.3% | Self-reported | Falcon2-11B Technical Report via Epoch AI | — |
| text-davinci-003 | 85.5% | Third party | The Falcon Series of Open Language Models via Epoch AI | — |
| text-davinci-003 | 85.5% | Self-reported | GPT-4 Technical Report via Epoch AI | — |
| DeepSeek-V2 | 87.1% | Self-reported | DeepSeek-V3 Technical Report via Epoch AI | — |
| DeepSeek-V3 | 88.9% | Self-reported | DeepSeek-V3 Technical Report via Epoch AI | — |
| Falcon-180B Chat | 89.0% | Self-reported | The Falcon Series of Open Language Models via Epoch AI | — |
| Llama 3.1 405B | 89.2% | Self-reported | DeepSeek-V3 Technical Report via Epoch AI | — |
| GPT-4 | 95.3% | Self-reported | GPT-4 Technical Report via Epoch AI | — |
| GPT-4 32K | 95.3% | Third party | The Falcon Series of Open Language Models via Epoch AI | — |