Model components / Fine-tuned / chat models

Command R+

Cohere's enterprise foundation model line, currently anchored by Command A+ (released May 20, 2026), a 218B-parameter sparse MoE (25B active) with native citation generation, W4A4 lossless quantization, and the company's first fully Apache 2.0 licensed frontier model. Runs on a single B200 or two H100s. Positioned for sovereign AI and enterprise RAG with native support for 48 languages and explicit grounding spans linking every factual claim to source documents. Earlier Command R+ (104B, 2024) is the predecessor; with Command A+ the line has flipped from API-only to open-weights.

Cohere c4ai-command-r-plus: 104B instruct/chat model, 128K context, RAG + multi-step tool use, 10 languages. Original release Apr 2024 (newer 08-2024 revision exists). Verified live on HF June 2026 (gated, contact-info required). License CC-BY-NC (non-commercial, non-OSI). Correctly an instruct model, belongs in finetuned_chat (was miscategorized as base elsewhere).

Openness

2 high confidence

2.0

weights: open(gated, contact-info required)
data: closed
code: closed
license: CC-BY-NC(non-commercial, non-OSI)+Cohere-Acceptable-Use-Policy

Weights downloadable but under CC-BY-NC, a non-commercial, non-OSI license with field-of-use restrictions => restricted. Marketed as 'open weights' despite the non-commercial cap.

https://huggingface.co/CohereLabs/c4ai-command-r-plus recorded 2026-06-04
License CC-BY-NC + Acceptable Use Policy, gated access, 104B instruct model

Adoption

1 high confidence

1.0

~3,258 HF downloads/mo on the original checkpoint (gated + superseded by 08-2024 revision + newer Command models). Non-commercial license and 104B size cap self-host adoption; some legacy use via Cohere API. Hobbyist/legacy scale on the download signal.

https://huggingface.co/CohereLabs/c4ai-command-r-plus recorded 2026-06-04
3,258 downloads/month, gated, newer revision available

Capability

2 medium confidence

2.0

A 2024-era frontier-ish chat/RAG model, now well below the 2026 open instruct frontier (no modern reasoning/agentic coding scores; legacy MMLU-class benchmarks only). Scored 2 against a 2026 comparison set despite being capable for its time.

https://huggingface.co/CohereLabs/c4ai-command-r-plus recorded 2026-06-04
Open LLM Leaderboard avg 74.6 with per-benchmark breakdown

Unchanged since 2026-06-24 (last edited, not re-checked)