Command R, Command R+, and Llama 3.1 70B Instruct all operate in the 35–104B parameter range and each targets retrieval-augmented generation and tool-use workloads. The comparison is really about two Cohere models with a specialized architecture versus Meta's general-purpose 70B with much broader hosting flexibility. Command R is Cohere's 35B-parameter open-weights model, released March 2024 with a 131K context window. Its architectural priorities — retrieval augmentation, structured tool-calling, multi-turn coherence — are consistent with its larger sibling Command R+. The 35B size typically translates to two-to-three times lower per-token cost than Command R+, making it the right starting point before evaluating whether the larger model's capability gain justifies the cost increase. The CC-BY-NC license applies: commercial production deployments should route through Cohere's own API rather than third-party hosts. Command R+ is Cohere's 104B flagship, released April 2024 with 131K context. It scores well on Cohere's own RAG evaluation suite and function-calling benchmarks, reflecting genuine architectural investment in grounding workflows. At 104B the per-token cost is substantially higher than the 35B version or Llama 3.1 70B. The CC-BY-NC license constraint means teams building commercial products should plan to use Cohere's first-party API. Meta's Llama 3.1 70B Instruct is a 70-billion-parameter model released July 2024 under the Llama 3 community license, with a 131K context window and MMLU scores around 79–80. It is the general-purpose alternative in this comparison — not specifically tuned for RAG, but good at it, and available across nearly every major inference provider at competitive rates. For teams that do not want to route through a single vendor's API, Llama 3.1 70B is the safer infrastructure choice. Pick Command R as the starting evaluation point for Cohere's RAG ecosystem. Pick Command R+ when the benchmarking shows the quality delta justifies 104B costs. Pick Llama 3.1 70B for production workloads where broad provider choice and the Llama community license are the deciding factors.