Math benchmark

# GSM8K leaderboard

> GSM8K results for 38 AI models, led by Deepseek Coder v2 at 94.5%. What the benchmark measures, who runs it, and a source for every score.
- Canonical page: https://noometry.com/benchmarks/gsm8k
- Last updated: 2026-10-10
- Title: GSM8K Leaderboard (October 2026): Scores by Model | Noometry

As of October 2026, Deepseek Coder v2 has the highest published GSM8K score on Noometry at 94.5%, out of 38 models with results.

Last verified October 10, 2026

## About GSM8K

Grade-school math word problems that need a few steps of arithmetic.

- **Category:** [Math](https://noometry.com/best/math)
- **Introduced:** 2021
- **Size:** 8,500 problems
- **Format:** Exact answer
- **Unit:** Percent (random guessing ≈ 0%)
- **Official site:** [arxiv.org](https://arxiv.org/abs/2110.14168)

## Top 15 models

Top models on GSM8K

1.  Deepseek Coder v2 94.5%
2.  Qwen2.5-Coder-32B 93%
3.  GPT-4 92%
4.  GPT-4o mini 91.3%
5.  Phi-3.5-MoE 88.7%
6.  Claude Instant 86.7%
7.  Qwen2.5-Coder (1.5B) 86.7%
8.  Phi-3.5-mini 86.2%
9.  Gemma 2 9B 84.9%
10.  Mistral Nemo 84.2%
11.  Gemini 1.5 Flash (May 2024) 82.4%
12.  Llama 3.1-8B 82.4%
13.  Yi-34B 76%
14.  Mixtral 8x7B 74.4%
15.  Llama 2-70B 69.6%
16.  60708090100

Sponsored placements are available on pages like this one. [Advertise on Noometry](https://noometry.com/advertise)

## All results

GSM8K results by model
| # | Model | Provider | Score | Setting | Source | Date |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | [Deepseek Coder v2](https://noometry.com/models/deepseek-coder-v2) |  [![](/logos/deepseek.svg) DeepSeek](https://noometry.com/providers/deepseek) | 94.5% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 2 | [Qwen2.5-Coder-32B](https://noometry.com/models/qwen2-5-coder-32b) |  [![](/logos/alibaba.svg) Alibaba (Qwen)](https://noometry.com/providers/alibaba) | 93% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 3 | [GPT-4](https://noometry.com/models/gpt-4) | [OpenAI](https://noometry.com/providers/openai) | 92% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 4 | [GPT-4o mini](https://noometry.com/models/gpt-4o-mini) | [OpenAI](https://noometry.com/providers/openai) | 91.3% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 5 | [Phi-3.5-MoE](https://noometry.com/models/phi-3-5-moe) |  [![](/logos/microsoft.svg) Microsoft](https://noometry.com/providers/microsoft) | 88.7% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 6 | [Claude Instant](https://noometry.com/models/claude-instant) | [Anthropic](https://noometry.com/providers/anthropic) | 86.7% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 7 | [Qwen2.5-Coder (1.5B)](https://noometry.com/models/qwen2-5-coder) |  [![](/logos/alibaba.svg) Alibaba (Qwen)](https://noometry.com/providers/alibaba) | 86.7% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 8 | [Phi-3.5-mini](https://noometry.com/models/phi-3-5-mini) |  [![](/logos/microsoft.svg) Microsoft](https://noometry.com/providers/microsoft) | 86.2% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 9 | [Gemma 2 9B](https://noometry.com/models/gemma-2-9b) |  [![](/logos/google.svg) Google](https://noometry.com/providers/google) | 84.9% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 10 | [Mistral Nemo](https://noometry.com/models/mistral-nemo) |  [![](/logos/mistral.svg) Mistral AI](https://noometry.com/providers/mistral) | 84.2% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 11 | [Gemini 1.5 Flash (May 2024)](https://noometry.com/models/gemini-1-5-flash) |  [![](/logos/google.svg) Google](https://noometry.com/providers/google) | 82.4% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 12 | [Llama 3.1-8B](https://noometry.com/models/llama-3-1-8b) |  [![](/logos/meta.svg) Meta](https://noometry.com/providers/meta) | 82.4% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 13 | [Yi-34B](https://noometry.com/models/yi-34b) | [01.AI](https://noometry.com/providers/01-ai) | 76% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 14 | [Mixtral 8x7B](https://noometry.com/models/mixtral-8x7b) |  [![](/logos/mistral.svg) Mistral AI](https://noometry.com/providers/mistral) | 74.4% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 15 | [Llama 2-70B](https://noometry.com/models/llama-2-70b) |  [![](/logos/meta.svg) Meta](https://noometry.com/providers/meta) | 69.6% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 16 | [Qwen-14B](https://noometry.com/models/qwen-14b) |  [![](/logos/alibaba.svg) Alibaba (Qwen)](https://noometry.com/providers/alibaba) | 61.3% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 17 | [GPT-3.5-turbo](https://noometry.com/models/gpt-3-5-turbo) | [OpenAI](https://noometry.com/providers/openai) | 57.8% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 18 | [StarCoder 2 15B](https://noometry.com/models/starcoder-2-15b) |  [![](/logos/nvidia.svg) NVIDIA](https://noometry.com/providers/nvidia) | 57.7% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 19 | [Falcon-180B](https://noometry.com/models/falcon-180b) |  [![](/logos/tii.svg) Technology Innovation Institute](https://noometry.com/providers/tii) | 54.4% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 20 | [Mistral 7B](https://noometry.com/models/mistral-7b) |  [![](/logos/mistral.svg) Mistral AI](https://noometry.com/providers/mistral) | 54.4% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 21 | [Falcon 2 11B](https://noometry.com/models/falcon-2-11b) |  [![](/logos/tii.svg) Technology Innovation Institute](https://noometry.com/providers/tii) | 53.8% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 22 | [Qwen-7B](https://noometry.com/models/qwen-7b) |  [![](/logos/alibaba.svg) Alibaba (Qwen)](https://noometry.com/providers/alibaba) | 51.7% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 23 | [Gemma 7B](https://noometry.com/models/gemma-7b) |  [![](/logos/google.svg) Google](https://noometry.com/providers/google) | 46.4% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 24 | [Nemotron-4 15B](https://noometry.com/models/nemotron-4-15b) |  [![](/logos/nvidia.svg) NVIDIA](https://noometry.com/providers/nvidia) | 46% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 25 | [Yi 6B](https://noometry.com/models/yi-6b) | [01.AI](https://noometry.com/providers/01-ai) | 44.9% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 26 | [Llama 2-34B](https://noometry.com/models/llama-2-34b) |  [![](/logos/meta.svg) Meta](https://noometry.com/providers/meta) | 42.2% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 27 | [INTELLECT-1](https://noometry.com/models/intellect-1) |  [![](/logos/huggingface.svg) Hugging Face](https://noometry.com/providers/huggingface) | 38.6% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 28 | [Llama 2-13B](https://noometry.com/models/llama-2-13b) |  [![](/logos/meta.svg) Meta](https://noometry.com/providers/meta) | 36.9% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 29 | [DeepSeek Coder 33B](https://noometry.com/models/deepseek-coder-33b) |  [![](/logos/deepseek.svg) DeepSeek](https://noometry.com/providers/deepseek) | 35.4% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 30 | [Falcon-40B](https://noometry.com/models/falcon-40b) |  [![](/logos/tii.svg) Technology Innovation Institute](https://noometry.com/providers/tii) | 33.8% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 31 | [StarCoder 2 7B](https://noometry.com/models/starcoder-2-7b) |  [![](/logos/nvidia.svg) NVIDIA](https://noometry.com/providers/nvidia) | 32.7% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 32 | [StarCoder 2 3B](https://noometry.com/models/starcoder-2-3b) |  [![](/logos/nvidia.svg) NVIDIA](https://noometry.com/providers/nvidia) | 21.6% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 33 | [DeepSeek Coder 6.7B](https://noometry.com/models/deepseek-coder-6-7b) |  [![](/logos/deepseek.svg) DeepSeek](https://noometry.com/providers/deepseek) | 21.3% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 34 | [Llama 13b](https://noometry.com/models/llama-13b) |  [![](/logos/meta.svg) Meta](https://noometry.com/providers/meta) | 20.6% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 35 | [Gemma 2B](https://noometry.com/models/gemma-2b) |  [![](/logos/google.svg) Google](https://noometry.com/providers/google) | 17.7% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 36 | [Llama 2-7B](https://noometry.com/models/llama-2-7b) |  [![](/logos/meta.svg) Meta](https://noometry.com/providers/meta) | 16.7% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 37 | [Falcon-7B](https://noometry.com/models/falcon-7b) |  [![](/logos/tii.svg) Technology Innovation Institute](https://noometry.com/providers/tii) | 6.8% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |
| 38 | [DeepSeek Coder 1.3B](https://noometry.com/models/deepseek-coder-1-3b) |  [![](/logos/deepseek.svg) DeepSeek](https://noometry.com/providers/deepseek) | 4.4% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |

## Compare the leaders

-   [Deepseek Coder v2 vs Qwen2.5-Coder-32B](https://noometry.com/compare/deepseek-coder-v2-vs-qwen2-5-coder-32b)
-   [Deepseek Coder v2 vs GPT-4](https://noometry.com/compare/deepseek-coder-v2-vs-gpt-4)
-   [Deepseek Coder v2 vs GPT-4o mini](https://noometry.com/compare/deepseek-coder-v2-vs-gpt-4o-mini)
-   [Deepseek Coder v2 vs Phi-3.5-MoE](https://noometry.com/compare/deepseek-coder-v2-vs-phi-3-5-moe)
-   [Qwen2.5-Coder-32B vs GPT-4](https://noometry.com/compare/gpt-4-vs-qwen2-5-coder-32b)
-   [Qwen2.5-Coder-32B vs GPT-4o mini](https://noometry.com/compare/gpt-4o-mini-vs-qwen2-5-coder-32b)

## Other math benchmarks

-   [FrontierMath (Tiers 1-3)](https://noometry.com/benchmarks/frontiermath)
-   [FrontierMath Tier 4](https://noometry.com/benchmarks/frontiermath-tier-4)
-   [MathArena Final-Answer Competitions](https://noometry.com/benchmarks/matharena)
-   [OTIS Mock AIME 2024-2025](https://noometry.com/benchmarks/otis-mock-aime)
-   [ProofBench](https://noometry.com/benchmarks/proofbench)
-   [Omni-MATH](https://noometry.com/benchmarks/omni-math)
-   [LMArena Math](https://noometry.com/benchmarks/arena-math)
-   [LiveBench Math](https://noometry.com/benchmarks/livebench-math)
-   [MATH Level 5](https://noometry.com/benchmarks/math-level-5)
-   [FrontierMath (Feb 2025 set)](https://noometry.com/benchmarks/frontiermath-2025-02) (reference)
-   [FrontierMath Erdős](https://noometry.com/benchmarks/frontiermath-erdos) (reference)
-   [FrontierMath Tier 4 (v1)](https://noometry.com/benchmarks/frontiermath-tier-4-v1) (reference)

## Frequently asked questions

### What does GSM8K measure?

Grade-school math word problems that need a few steps of arithmetic.

### Which model has the highest GSM8K score?

As of October 2026, Deepseek Coder v2 has the highest published GSM8K score on Noometry at 94.5%, out of 38 models with results.

### What is the best open-weight model on GSM8K?

Deepseek Coder v2 has the highest GSM8K accuracy among open-weight models at 94.5%, ranking 1 of 38 overall.

### Cite this page

Noometry. (2026). GSM8K leaderboard. Retrieved October 10, 2026, from https://noometry.com/benchmarks/gsm8k

Quote Noometry with a link back to this page. It is also available in [Markdown](https://noometry.com/md/benchmarks/gsm8k.md).