HellaSwag leaderboard
As of 6 Nov 2026, the top score on HellaSwag is 95.3% by GPT-4 (OpenAI), ahead of Llama 3.1 405B at 89.2%, across 68 model configurations with a published result (Epoch AI Benchmarking Hub, result dated 14 Mar 2023). Commonsense sentence completion. HellaSwag is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 6 Nov 2026)
- GPT-4 holds the top HellaSwag score at 95.3% (Epoch AI Benchmarking Hub, 14 Mar 2023).
- Llama 3.1 405B is second on HellaSwag at 89.2% (Epoch AI Benchmarking Hub, 23 Jul 2024).
- HellaSwag: Commonsense sentence completion. A 2019 benchmark, saturated for current models; kept for history (Epoch AI Benchmarking Hub).
- 68 model configurations have a published HellaSwag result on BenchLeader (BenchLeader, 6 Nov 2026).
- HellaSwag does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
Commonsense sentence completion. A 2019 benchmark, saturated for current models; kept for history.
- Published by
- Epoch AI Benchmarking Hub
- Category
- Reasoning
- Index weight
- Reference only
- Models
- 68
- Data as of
- 6 Nov 2026
CC BY 4.0 — Epoch AI, ‘AI Benchmarking Hub’, epoch.ai/benchmarks. Mirrored boards credit their original publishers.
What the test looks like
Commonsense sentence completion.
How it is scored
Accuracy, as compiled by Epoch AI from published results.
What to keep in mind
Saturated: frontier models score near the ceiling, so it separates only older or smaller models, and results are largely self-reported by developers.
- 1GPT-495.3%
- 2Llama 3.1 405B89.2%
- 3Falcon-180B89.0%
- 4DeepSeek V388.9%
- 5Deepseek v287.1%
- 6PaLM 2-L86.8%
- 7Mixtral 8x7B86.7%
- 8davinci-00385.5%
- 9Llama 2-70B85.3%
- 10Falcon-40B85.3%
- 11Qwen2.5 72B84.8%
- 12LLaMA-65B84.2%
- 13Stable Beluga 284.1%
- 14PaLM 2-M84.0%
- 15PaLM 540B83.8%
68 of 68
| # | ||||
|---|---|---|---|---|
| 1 |  GPT-4 | 95.3% | 42.4 | 2023-03-14 |
| 2 |  Llama 3.1 405Bopen ↗ | 89.2% | 41.9 | 2024-07-23 |
| 3 |  Falcon-180B | 89.0% | – | 2023-09-06 |
| 4 |  DeepSeek V3open ↗ | 88.9% | 44.4 | 2024-12-26 |
| 5 | Deepseek v2open ↗ | 87.1% | – | 2024-05-07 |
| 6 |  PaLM 2-L | 86.8% | – | 2023-05-17 |
| 7 |  Mixtral 8x7Bopen | 86.7% | 30.9 | 2023-12-11 |
| 8 | davinci-003 | 85.5% | – | 2022-11-28 |
| 9 | Llama 2-70Bopen ↗ | 85.3% | 33.6 | 2023-07-18 |
| 10 | Falcon-40B | 85.3% | – | 2023-03-15 |
| 11 |  Qwen2.5 72Bopen ↗ | 84.8% | 43.2 | 2024-09-19 |
| 12 | LLaMA-65Bopen | 84.2% | – | 2023-02-24 |
| 13 |  Stable Beluga 2open | 84.1% | – | 2023-07-20 |
| 14 | PaLM 2-M | 84.0% | – | 2023-05-17 |
| 15 | PaLM 540B | 83.8% | – | 2022-04-04 |
| 16 | Qwen2.5 Coder 32B Instructopen ↗ | 83.0% | 42.7 | 2024-09-18 |
| 17 | Falcon 2 11Bopen | 82.9% | – | 2024-05-09 |
| 18 | LLaMA-33B | 82.8% | – | 2023-02-24 |
| 19 |  Phi 3medium | 82.4% | 37.9 | 2024-04-23 |
| 20 |  Megatron-Turing NLG 530B | 82.4% | – | 2021-10-11 |
| 21 | Nemotron-4 15B | 82.4% | – | 2024-02-26 |
| 22 | Gemma 7B | 82.2% | – | 2024-02-21 |
| 23 | PaLM 2-S | 82.0% | – | 2023-05-17 |
| 24 | text-davinci-002 | 81.5% | – | 2022-03-15 |
| 25 | Mistral 7B v0.3open | 81.0% | 29.0 | 2023-09-27 |
| 26 | Llama 2-13Bopen ↗ | 80.7% | – | 2023-07-18 |
| 27 | Qwen2.5-Coder-14B | 80.2% | – | 2024-09-18 |
| 28 | InstructGPT 175B | 79.3% | – | 2022-01-27 |
| 29 | LLaMA-13B | 79.2% | – | 2023-02-24 |
| 30 | OPT-175B | 79.1% | – | 2022-05-02 |
| 31 | Falcon-7B | 78.1% | – | 2023-04-24 |
| 32 |  InternLM 20B | 78.1% | – | 2023-09-18 |
| 33 | Llama 2-7Bopen ↗ | 77.2% | – | 2023-07-18 |
| 34 | Phi 3 Small | 77.0% | – | 2024-04-23 |
| 35 | Qwen2.5-Coder 7B Instructopen ↗ | 76.8% | – | 2024-09-18 |
| 36 | Phi 3 Miniopen | 76.7% | 36.9 | 2024-04-23 |
| 37 |  MPT-7B | 76.4% | – | 2023-05-05 |
| 38 |  Yi-9B | 76.4% | – | 2024-03-01 |
| 39 | LLaMA-7B | 76.2% | – | 2023-02-24 |
| 40 | OPT-66B | 74.5% | – | 2022-05-03 |
| 41 | Yi 6B | 74.4% | – | 2023-11-02 |
| 42 |  Xgen 7B Base | 74.2% | – | 2023-06-27 |
| 43 | open_llama_7b | 71.8% | – | 2023-06-07 |
| 44 |  INTELLECT-1 | 71.4% | – | 2024-11-29 |
| 45 | Gemma 2B | 71.4% | – | 2024-02-21 |
| 46 | Qwen2.5-Coder-3B | 70.9% | – | 2024-09-18 |
| 47 |  Baichuan2-13Bopen | 70.8% | – | 2023-09-06 |
| 48 |  Dolly v2 12B | 70.8% | – | 2023-04-11 |
| 49 | InternLM 7B | 70.6% | – | 2023-07-05 |
| 50 |  GPT-NeoX-20B | 70.5% | – | 2022-04-07 |
| 51 |  RedPajama-INCITE-7B-Base | 70.3% | – | 2023-05-04 |
| 52 | OPT-13B | 69.9% | – | 2022-05-11 |
| 53 | Baichuan 2 7B Base | 68.0% | – | 2023-09-20 |
| 54 | InstructGPT 6B | 67.6% | – | – |
| 55 | GPT-J-6B | 66.2% | – | 2021-08-05 |
| 56 | Qwen2.5 Coder 1.5B | 61.8% | – | 2024-09-18 |
| 57 |  Cerebras-GPT-13B | 59.4% | – | 2023-03-20 |
| 58 |  Vicuna 13B v1.1 | 57.8% | – | 2023-04-12 |
| 59 |  ChatGLM2-6B | 57.0% | – | 2023-06-24 |
| 60 | InstructGPT 1.3B | 56.1% | – | – |
More reasoning benchmarks · All LLM benchmarks
Related benchmarks
Cite as: BenchLeader, “HellaSwag leaderboard”, data as of 6 Nov 2026, https://www.benchleader.com/benchmarks/hellaswag.
HellaSwag: questions
- What does HellaSwag measure?
- Commonsense sentence completion. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on HellaSwag?
- As of 6 Nov 2026, the highest score on HellaSwag is 95.3% by GPT-4 (result dated 14 Mar 2023), ahead of Llama 3.1 405B at 89.2%. Scores come from Epoch AI Benchmarking Hub and are re-read daily.
- Which AI model leads HellaSwag?
- GPT-4 leads HellaSwag with 95.3% as of 6 Nov 2026 (result dated 14 Mar 2023), ahead of Llama 3.1 405B at 89.2%.
- How many models have HellaSwag results?
- 68 model configurations have a HellaSwag result on BenchLeader, all taken from Epoch AI Benchmarking Hub.
- Who runs HellaSwag and how often is it updated?
- HellaSwag is published by Epoch AI Benchmarking Hub. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does HellaSwag count toward the BenchLeader Index?
- No. HellaSwag is shown for reference but left out of the composite index.