# deepseek-ai/DeepSeek-Math

DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models

Repository: https://github.com/deepseek-ai/DeepSeek-Math
Canonical: https://ross.abutalabs.com/products/deepseek-math
Language: Python
License: MIT
License Family: permissive
Last push: 2024-04-15T07:55:37+00:00

## Health v2 (maintenance only)
Score: 26/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 67
- inputs: {"age_days": 940, "days_push": 870, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3436, forks 593 (observed 2026-08-28T04:08:04.950228+00:00)

## What it is
DeepSeekMath is a 7B open language model specialized in mathematical reasoning, initialized from DeepSeek-Coder and trained on 500B math-related tokens. The repository provides model checkpoints (base, instruct, RL), inference code, and evaluation tooling for the MATH benchmark.

## Use cases
- solve competition-level math problems with an LLM
- run a 7B math reasoning model locally
- fine-tune a model for mathematical question answering
- evaluate language models on the MATH benchmark
- compare open math models against GPT-4 performance

## When to choose
- you need an open-weight model strong at mathematical reasoning
- you want base, instruct, and RL checkpoints for research
- you need a small model that scores well on MATH without external tools

## When to avoid
- you need a general-purpose assistant rather than math specialization
- you lack GPU resources for 7B model inference
- you need a maintained product with ongoing updates

## Facets
- artifact type: library
- maturity: maintenance
- function: llm-inference, llm-training, machine-learning
- domain: large-language-models, mathematics, deep-learning
- platform: python
- tags: math-reasoning, open-language-model, model-weights, huggingface, reinforcement-learning, natural-language-processing, gpu, linux

## Member repositories
- deepseek-ai/DeepSeek-Math (main) score 26

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:04.950228+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:37:42.006236+00:00, confidence not recorded.
  - readme: https://github.com/deepseek-ai/DeepSeek-Math (fetched 2026-08-28T04:08:04.950228+00:00, sha d56fd7dc1fd0)
- Data as of 2026-08-30T08:39:29.467469+00:00.