The Open-Source AI Stack
RSS
All models

Models · command

Command A+

Open Cohere · 2026-05-20 · Apache-2.0

Cohere's first full Apache 2.0 model, released May 20 2026 as a 218B MoE activating 25B per token. Aimed at sovereign critical infrastructure deployments; fine-tuneable on classified data and runnable air-gapped. Distributed in BF16, FP8, and a W4A4 4-bit format. Multimodal and multilingual across 48 languages.

Cost

$0.00 / Mtok input
$0.00 / Mtok output

· as of 2026-05-21

source ↗

Speed

212.2 tok/sec output
157 ms TTFT

· as of 2026-05-21

source ↗

Architecture

tokens in Embedding vocab not disclosed × N layers Attention (not disclosed) Position encoding not disclosed context 128,000 tokens MoE Router ? experts total · ? active per token Output projection tokens out
Schema-generated from data/models.yaml. Every label is auditable against the model's sources.

Specs

Architecture
moe
Total params
218B
Active params
25B
Context window
128K tokens
Attention
unknown
Position encoding
unknown
Post-training
sft, rlhf
OSI-approved
yes
Data released
no
Training code
not released

Benchmarks

Each score carries the date it was published; we never infer or interpolate missing scores.

General reasoning

GPQA-Diamond 76.1 as of 2026-05-21 source ↗

Available quantizations

GGUF llama.cpp's container; the common local format, k-quants from Q2 to Q8. runs on llama.cpp, Ollama
FP8 8-bit float, frequently a native release on Hopper / Blackwell GPUs. runs on vLLM, SGLang, TensorRT-LLM

Verified via the Hugging Face model tree ↗. Community quantizations change over time; the families shown are those with published weights at audit time.

Notable innovations

  • · First Apache 2.0 Cohere model
  • · W4A4 4-bit native distribution
  • · Native citations via lossless quantization claim
  • · Sovereign-deployment focus (air-gapped, on-prem)

Lineage

First Apache 2.0 Cohere model; targets sovereign / air-gapped enterprise.

Derived from

Command A 2025-03-13

Sources