Loading...
OpenAI · DeepSeek — Interactive pricing comparison
DeepSeek R1 outscores o4 Mini on reasoning (72 vs 68) at half the price on both input and output. o4 Mini buys you vision and function-calling that R1 lacks entirely.
OpenAI
DeepSeek
DeepSeek R1 is 85% cheaper at this usage
o4 Mini has a larger context window
The model that calls itself a reasoning specialist loses the reasoning benchmark to the cheaper competitor. o4 Mini scores 68 reasoning. DeepSeek R1 scores 72. On the dimension that defines both models, R1 wins by a margin that matters.
Pricing reinforces the gap. o4 Mini charges $1.10 input and $4.40 output. R1 charges $0.55 and $2.19. That is a 2x spread on both ends. For any pipeline that generates long chain-of-thought traces — which reasoning models do by design — the output difference compounds fast. A single reasoning run that emits 10K tokens costs $0.044 on o4 Mini versus $0.022 on R1.
But o4 Mini has capabilities R1 simply does not. Vision lets it read screenshots, diagrams, and charts. Function-calling lets it drive tools, query APIs, and orchestrate multi-step workflows. R1 has neither. It is a text-in, text-out reasoner with no ability to interact with the outside world.
Context is close enough to not matter for most reasoning tasks: 200K for o4 Mini, 128K for R1. Both exceed what typical chain-of-thought problems require.
R1 if your reasoning runs on text alone and volume matters. o4 Mini if your reasoning needs to see images or call tools. The reasoning scores say R1; the capability set says o4 Mini. Most teams that start with o4 Mini for its tool integration end up routing their pure-reasoning tasks to R1 once the output bill arrives.
Last updated June 2026