GSM8K
Grade School Math 8K - grade-school math word problems. 8-shot.
Category
Math
Score type
percent
Scale
100%
higher is better
Models tested
6
What is GSM8K?
GSM8K contains 8,000 grade-school math word problems. They look simple but require several correct steps in a row. The benchmark has become the standard test of basic math reliability; today's frontier models score high, which is why it mainly separates smaller and older models.
What does it test?
GSM8K measures mathematical problem solving. In detail, it checks:
- reading comprehension leading to an equation setup
- stepwise arithmetic executed correctly
- a single correct final numeric answer
- consistent performance across many short problems
How to read the score
The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.
Frequently asked questions
What is the GSM8K benchmark?
GSM8K contains 8,000 grade-school math word problems. They look simple but require several correct steps in a row. The benchmark has become the standard test of basic math reliability; today's frontier models score high, which is why it mainly separates smaller and older models.
What does GSM8K test?
GSM8K mainly measures mathematical problem solving. Concretely, it checks: reading comprehension leading to an equation setup; stepwise arithmetic executed correctly; a single correct final numeric answer.
How is the GSM8K score calculated?
The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.
Where do I get GSM8K data and scores?
The official source is https://arxiv.org/abs/2110.14168. The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/gsm8k/scores) or pull it directly from the Open LLM Radar API with any HTTP client.
Benchmark data & downloads
Where to get, inspect and reuse the data behind this benchmark.
Official benchmark page / paper
The benchmark maintainer's website, paper or leaderboard
Benchmark dataset
The question/task data itself — download and evaluate locally
Benchmark metadata (JSON)
This page's benchmark record from the Open LLM Radar API
All model scores (JSON download)
Full score table as JSON; paginate with limit/offset params
Open LLM Radar API docs
Swagger UI to explore and reuse all radar API endpoints
Model scores 6 models tested
| # | Model | Score | Date | Source | Verified |
|---|---|---|---|---|---|
| 1 |
Bonsai 2 27B — GGUF
Prism ML
|
96.66%
|
17.09.2026 | Report | Verified |
| 2 |
Gemma 3 27B PT
Google DeepMind
|
82.6%
|
27.02.2025 | Report | Verified |
| 3 |
Clef
Cloudflare
|
80.8%
|
— | Report | Verified |
| 4 |
Gemma 3 12B PT
Google DeepMind
|
71%
|
27.02.2025 | Report | Verified |
| 5 |
Clef-Flash
Cloudflare
|
67.3%
|
— | Report | Verified |
| 6 |
Gemma 3 4B PT
Google DeepMind
|
38.4%
|
27.02.2025 | Report | Verified |