DeepSeek R1 0528 (May '25) vs GPT-4o (Nov '24)

Intelligence Index

MMLU-Pro

GPQA

AIME

Which is cheaper, DeepSeek R1 0528 (May '25) or GPT-4o (Nov '24)?

DeepSeek R1 0528 (May '25) is cheaper on list price. DeepSeek R1 0528 (May '25) costs $1.35/M input and $4.2/M output tokens. GPT-4o (Nov '24) costs $2.5/M input and $10.0/M output tokens. On combined list price, DeepSeek R1 0528 (May '25) is 2.3x cheaper than GPT-4o (Nov '24). However, list prices alone can be misleading because different models use different numbers of tokens for the same task. Check the effective cost comparison above, which adjusts for token efficiency using benchmark data.

Which scores higher on benchmarks, DeepSeek R1 0528 (May '25) or GPT-4o (Nov '24)?

DeepSeek R1 0528 (May '25) has a higher Intelligence Index (27.0) compared to GPT-4o (Nov '24) (17.3). The Intelligence Index is a composite score from three industry-standard benchmarks: MMLU-Pro (general knowledge and reasoning), GPQA (graduate-level science), and AIME (mathematical problem solving). A higher score means the model produces more accurate and capable responses across a broad range of tasks. This composite approach is more reliable than any single benchmark because it measures different types of capability.

Which model is better value for money, DeepSeek R1 0528 (May '25) or GPT-4o (Nov '24)?

DeepSeek R1 0528 (May '25) offers better value at $0.21 per intelligence point compared to GPT-4o (Nov '24) at $0.72 per intelligence point. Cost per intelligence point measures how much you pay for each unit of benchmark performance, calculated as the combined token cost divided by the Intelligence Index score. When token efficiency data is available, this calculation uses effective prices (adjusted for the fact that different models consume different numbers of tokens for the same task) rather than raw list prices. A lower cost per intelligence point means you get more capability per dollar.

Which has a larger context window, DeepSeek R1 0528 (May '25) or GPT-4o (Nov '24)?

DeepSeek R1 0528 (May '25) supports 128,000 tokens compared to GPT-4o (Nov '24) with 128,000 tokens. The context window determines how much text (including your prompt, conversation history, and documents) the model can process in a single request. A larger context window is important for tasks like document summarization, long-form analysis, and multi-turn conversations with extensive history. If your use case involves processing large inputs, the context window may be a deciding factor.