GSM8K leaderboard
As of 6 Nov 2026, the top score on GSM8K is 94.5% by Deepseek Coder v2 (DeepSeek), ahead of Qwen2.5-Coder-14B at 94.2%, across 73 model configurations with a published result (Epoch AI Benchmarking Hub, result dated 17 Jun 2024). Grade-school maths word problems. GSM8K is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 6 Nov 2026)
- Deepseek Coder v2 holds the top GSM8K score at 94.5% (Epoch AI Benchmarking Hub, 17 Jun 2024).
- Qwen2.5-Coder-14B is second on GSM8K at 94.2% (Epoch AI Benchmarking Hub, 6 Nov 2024).
- GSM8K: Grade-school maths word problems. A 2021 benchmark, saturated for current models; kept for history (Epoch AI Benchmarking Hub).
- 73 model configurations have a published GSM8K result on BenchLeader (BenchLeader, 6 Nov 2026).
- GSM8K does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
Grade-school maths word problems. A 2021 benchmark, saturated for current models; kept for history.
- Published by
- Epoch AI Benchmarking Hub
- Category
- Maths
- Index weight
- Reference only
- Models
- 73
- 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
Grade-school maths word problems.
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.
- 1Deepseek Coder v294.5%
- 2Qwen2.5-Coder-14B94.2%
- 3Qwen2.5 Coder 32B Instruct93.0%
- 4GPT-492.0%
- 5GPT-4o mini91.3%
- 6GPT 4 061390.0%
- 7Phi-3.5-MoE88.7%
- 8DeepSeek-Coder-V2-Lite-Instruct87.6%
- 9Claude Instant 1.286.7%
- 10Qwen2.5-Coder 7B Instruct86.7%
- 11Phi-3.5-mini86.2%
- 12DeepSeek-Coder-V2 236B85.8%
- 13Gemma 2 9B84.9%
- 14Mistral Nemo Base 240784.2%
- 15Llama 3.1 8B82.4%
73 of 73
| # | ||||
|---|---|---|---|---|
| 1 |  Deepseek Coder v2open ↗ | 94.5% | – | 2024-06-17 |
| 2 |  Qwen2.5-Coder-14B | 94.2% | – | 2024-11-06 |
| 3 | Qwen2.5 Coder 32B Instructopen ↗ | 93.0% | 42.7 | 2024-11-21 |
| 4 |  GPT-4 | 92.0% | 42.4 | 2023-03-14 |
| 5 | GPT-4o mini | 91.3% | 35.4 | 2024-07-18 |
| 6 | GPT 4 0613 | 90.0% | 39.5 | 2023-06-13 |
| 7 |  Phi-3.5-MoE | 88.7% | – | 2024-08-17 |
| 8 | DeepSeek-Coder-V2-Lite-Instructopen ↗ | 87.6% | – | 2024-06-13 |
| 9 |  Claude Instant 1.2 | 86.7% | – | 2023-08-09 |
| 10 | Qwen2.5-Coder 7B Instructopen ↗ | 86.7% | – | 2024-09-17 |
| 11 | Phi-3.5-mini | 86.2% | – | 2024-08-16 |
| 12 | DeepSeek-Coder-V2 236B | 85.8% | – | 2024-06-17 |
| 13 |  Gemma 2 9B | 84.9% | 38.8 | – |
| 14 |  Mistral Nemo Base 2407 | 84.2% | – | 2024-07-18 |
| 15 |  Llama 3.1 8Bopen ↗ | 82.4% | 33.7 | 2024-07-23 |
| 16 | Gemini 1.5 Flash 001 | 82.4% | – | 2024-05-23 |
| 17 | Claude Instant 1.1 | 80.9% | – | – |
| 18 | Qwen2.5-Coder-3B | 80.7% | – | 2024-11-06 |
| 19 | davinci-003 | 78.2% | – | 2022-11-28 |
| 20 |  Yi-34B | 76.0% | 34.9 | 2023-11-22 |
| 21 | Mixtral 8x7Bopen | 74.4% | 30.9 | 2023-12-11 |
| 22 | Llama 2-70Bopen ↗ | 69.6% | 33.6 | 2023-07-18 |
| 23 |  Stable Beluga 2open | 69.6% | – | 2023-07-20 |
| 24 | DeepSeek-Coder-V2-Lite-Base | 67.1% | – | 2024-06-13 |
| 25 | Qwen2.5 Coder 1.5B | 65.8% | – | 2024-09-18 |
| 26 |  InternLM 20B | 62.9% | – | 2023-09-18 |
| 27 | Qwen-14B | 61.3% | – | 2023-09-28 |
| 28 | GPT-3.5 Turbo (older v0613) | 57.8% | – | 2023-06-13 |
| 29 |  StarCoder2-15Bopen | 57.7% | – | 2024-02-20 |
| 30 | code-davinci-002 | 56.8% | – | 2022-03-15 |
| 31 | PaLM 540B | 56.5% | – | 2022-04-04 |
| 32 | Mistral 7B v0.3open | 54.4% | 29.0 | 2023-09-27 |
| 33 |  Falcon-180B | 54.4% | – | 2023-09-06 |
| 34 | LLaMA-65Bopen | 54.4% | – | 2023-02-24 |
| 35 | Falcon 2 11Bopen | 53.8% | – | 2024-05-09 |
| 36 |  Baichuan2-13Bopen | 52.8% | – | 2023-09-06 |
| 37 | Qwen-7B | 51.7% | – | 2023-09-28 |
| 38 | Gemma 7B | 46.4% | – | 2024-02-21 |
| 39 |  Nemotron-4 15B | 46.0% | – | 2024-02-26 |
| 40 | Yi 6B | 44.9% | – | 2023-11-22 |
| 41 | LLaMA-33B | 44.1% | – | 2023-02-24 |
| 42 | Llama 2-34B | 42.2% | – | 2023-07-18 |
| 43 | text-davinci-002 | 41.5% | – | 2022-03-15 |
| 44 |  INTELLECT-1 | 38.6% | – | 2024-11-29 |
| 45 | CodeQwen1.5-7B | 37.7% | – | 2024-04-15 |
| 46 | Llama 2-13Bopen ↗ | 36.9% | – | 2023-07-18 |
| 47 | Deepseek Coder 33B Base | 35.4% | – | 2023-11-02 |
| 48 | Qwen2.5-Coder-0.5B | 34.5% | – | 2024-09-18 |
| 49 |  MPT-30B | 34.4% | – | 2023-06-22 |
| 50 | Falcon-40B | 33.8% | – | 2023-05-25 |
| 51 | PaLM 62B | 33.0% | – | 2022-04-04 |
| 52 | StarCoder2-7Bopen | 32.7% | – | 2024-02-20 |
| 53 |  ChatGLM2-6B | 32.4% | – | 2023-06-24 |
| 54 | InternLM 7B | 31.2% | – | 2023-07-05 |
| 55 |  Vicuna 13B v1.1 | 28.1% | – | 2023-04-12 |
| 56 | Baichuan 1-13B | 26.8% | – | 2023-07-11 |
| 57 | Baichuan 2 7B Base | 24.6% | – | 2023-09-20 |
| 58 | StarCoder2-3Bopen | 21.6% | – | 2024-02-22 |
| 59 | Deepseek Coder 6 7B Base | 21.3% | – | 2023-11-02 |
| 60 | Qwen-1_8B | 21.2% | – | 2023-11-30 |
More maths benchmarks · All LLM benchmarks
Cite as: BenchLeader, “GSM8K leaderboard”, data as of 6 Nov 2026, https://www.benchleader.com/benchmarks/gsm8k.
GSM8K: questions
- What does GSM8K measure?
- Grade-school maths word problems. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on GSM8K?
- As of 6 Nov 2026, the highest score on GSM8K is 94.5% by Deepseek Coder v2 (result dated 17 Jun 2024), ahead of Qwen2.5-Coder-14B at 94.2%. Scores come from Epoch AI Benchmarking Hub and are re-read daily.
- Which AI model leads GSM8K?
- Deepseek Coder v2 leads GSM8K with 94.5% as of 6 Nov 2026 (result dated 17 Jun 2024), ahead of Qwen2.5-Coder-14B at 94.2%.
- How many models have GSM8K results?
- 73 model configurations have a GSM8K result on BenchLeader, all taken from Epoch AI Benchmarking Hub.
- Who runs GSM8K and how often is it updated?
- GSM8K 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 GSM8K count toward the BenchLeader Index?
- No. GSM8K is shown for reference but left out of the composite index.