DeepSeek R1 0528, the DeepSeek model in numbers

DeepSeek R1 0528 at a glance

DeepSeek R1 0528 is a language model from DeepSeek, released on May 28, 2025. That makes it more than a year old: successors may have shipped since. It is an open-weight model: you can host it yourself on infrastructure you control, and 16 API providers listed on models.dev also offer it. It has a reasoning mode: it can "think" before answering, which helps with multi-step problems but uses more tokens, and therefore more money.

We file it under open-weight models (68 models in this category on the site).

Benchmark results

Epoch AI publishes 3 results for DeepSeek R1 0528 on our basket's benchmarks: too few to compute a reliable capability index.

• GPQA Diamond (PhD-level science questions): 76.3% — the best tracked model, Gemini 4 Argon, reaches 96.4%
• OTIS Mock AIME (competition mathematics problems): 66.4% — the best tracked model, GPT-5.5 Pro, reaches 100%
• Aider Polyglot (code editing across several languages): 7,140% — the best tracked model, GPT-5, reaches 8,800%

What it costs

DeepSeek R1 0528 costs $0.52 per million input tokens and $2.21 per million output tokens (median of 16 providers, checked on October 22, 2026).

In practice, an agent task that reads 20,000 tokens and writes 2,000 costs about $0.015, or $14.88 for 1,000 such tasks.

That is 2.2 times the median output price of open-weight models ($1).

To estimate your own budget, use the agent cost calculator.

What it can do

Its context window reaches 164,000 tokens, roughly 246 pages of text at 500 words per page: enough for a medium-length document. A single answer can run to 32,768 tokens.

It accepts as input: text.

It can call tools (functions, APIs) and return structured JSON output, which makes it a possible foundation for an AI agent.

The weights are published on Hugging Face (deepseek-ai/DeepSeek-R1-0528). See also our list of open-source models.

When to choose it

Given its category, DeepSeek R1 0528 is mainly aimed at projects where you want control over your data or hosting, or want to adapt the model to your field.

Before adopting it, test it on your own examples: benchmarks measure standard tasks, not your job.

Comparable models

In the same category, the models closest to DeepSeek R1 0528 are:

• GLM-4.7 (Zhipu AI (Z.ai), index 68.4)
• Nemotron 3 Ultra 550B A55B (NVIDIA, $2.4 per million output tokens)
• Devstral 2 (Mistral AI, $2 per million output tokens)
• DeepSeek V4 Pro 0813 (DeepSeek, index 72.2)