Within Cohere's own lineup, Command R+ is roughly 3–4× the price of Command R per token — an unusually large spread for an intra-family comparison. Command R sits around 35B parameters; Command R+ runs closer to 104B. Both share the same 128K context window and the same retrieval-grounding prompt format, so the migration path between them is clean. The benchmark gap is real but task-dependent. On MMLU, Command R+ scores roughly 75% versus Command R's ~68%. The difference is larger on multi-hop reasoning and complex instruction-following; for single-turn retrieval QA the gap narrows significantly. Cohere reports Command R+ performs notably better on tool-use and structured output tasks — relevant for agentic pipelines. Command R is the right pick for high-volume classification, sentiment analysis, or single-document summarization pipelines where you're pushing millions of requests daily. At roughly $0.15/1M input tokens on competitive providers, the math works out well for workloads where a smaller quality delta is acceptable. Review live pricing on [Command R's model page](/models/cohere--command-r). Command R+ earns its cost premium on multi-document synthesis, complex RAG with multiple retrieved passages, and multi-step tool-calling agents. If your pipeline involves chaining retrieval with tool calls and the model needs to track state across a long context, the capability lift from the larger model is often worth the 3–4× cost. See provider options on [Command R+'s model page](/models/cohere--command-r-plus). **Pick Command R** for volume-sensitive, lower-complexity retrieval tasks. **Pick Command R+** for agentic workflows, multi-hop reasoning, and complex tool use.