R1 0528
DeepSeek's mid-tier model. Context window: 164K tokens.
Scores by test
Methodology →What you need to know
R1 0528 is a high-performance model that excels in precision-oriented tasks, specifically agentic planning, tool calling, and faithfulness. Its strongest technical differentiator is its mathematical capability, evidenced by a 96.6% score on MATH Level 5 and 66.4% on AIME 2025. These metrics indicate a model capable of handling complex logical reasoning and quantitative analysis far beyond standard general-purpose models.
The model maintains a high internal quality average of 4.46/5.0, with perfect scores in persona consistency, multilingual support, and long context handling. With a 164K context window and top-tier scores in faithfulness, it is reliable for processing large datasets without losing coherence or introducing hallucinations.
At a blended cost of $1.74/MTok, the model is priced competitively for its rank (#12 of 71). It provides elite-level reasoning and agentic capabilities at a price point that is significantly lower than many proprietary models in the same performance tier.
Use this model if your application requires rigorous mathematical accuracy, complex tool integration, or stable persona management across long conversations. Skip this model if your primary need is highly creative writing or nuanced strategic analysis, where it scores slightly lower relative to its own technical benchmarks.
Strengths — Top 3
Relative weaknesses — Bottom 3
Similar models