Command models + North by Cohere: status and timeline

Command models + North

Provide efficient, deployable AI models and platforms that enterprises in regulated industries can afford to run and trust.

ScalingStage 5 of 5

Command A+ (218B sparse MoE) released May 2026. North enterprise platform in active deployment.

Updated May 2026Checked 11 Oct0 updates this week

Milestones

Next · IPO expected

Cohere founded by Google Brain researchers2019Complete.
Series B funding, $200M2023Complete.
Series C funding, $500M+2024Complete.
Command R+ released with 128K context2024Complete.
$240M ARR, 50%+ QoQ growth achieved2025Complete.
Command A+ sparse MoE releasedMay 2026Complete.
IPO expectedTarget 2026Not yet reached.

Most important updates

Upcoming

2026IPO expected (next)
2026IPO filing or announcement

Current obstacles

IPO execution in a crowded marketMultiple AI labs (OpenAI, Anthropic, xAI) are planning public offerings around the same time, and market appetite for AI company IPOs is uncertain.
Scale without losing efficiencyAs Command models grow larger (218B parameters), the advantage of efficiency per token erodes if inference cost scales faster than performance gains.

Physics limits

Model size and inference speed trade-offLarger models (Command A+ 218B) are more capable but slower; smaller models run fast but lose reasoning ability. This trade-off defines the Pareto frontier for any deployment.
Context window limits memoryLonger context windows (128K tokens) allow the model to 'read' more, but each added token increases memory use and latency quadratically in the attention mechanism.

How it works

3 parts

Efficient inference

Lower cost per token

Optimized inference lets enterprises run models on their own infrastructure without expensive GPU clusters.

RAG integration

Grounded answers with citations

Optimized for retrieval-augmented generation, pulling facts from company data and citing sources.

North workspace

Secure agent orchestration

Platform layer for building multi-step AI agents with security, audit logs and compliance hooks.

Spec sheet

Spec	Command models + North
2025 ARR	$240M^R (reported)
QoQ growth rate	50%+^R (reported)
Gross margin	~70%^E (estimated)
Command A+ parameters	218B (sparse MoE)^R (reported)
Command R+ context window	128,000 tokens^R (reported)

^R reported by the company · ^E our estimate

Papers & demos

1 item

2017paper
Attention Is All You Need (opens arxiv.org)Co-authored by Aidan Gomez; introduced the Transformer architecture underpinning all modern large language models.

Update log

2 updates

May 2026

Major: Cohere releases Command A+, a sparse mixture-of-experts model with 218B parametersBlogSoftware

2024

Major: Cohere raises $500M Series C led by Nvidia and SalesforcePressFunding

About Cohere

The team behind Command models + North

Cohere

Command models, North enterprise platform, AI for regulated industries

Cohere builds large language models and AI platforms optimized for enterprises in regulated industries: finance, healthcare, manufacturing, energy and government. Founded by Google Brain researchers, it reached $240M ARR in 2025 with 50%+ quarter-on-quarter growth and plans to IPO in 2026.

Cohere surpassed $240M ARR in 2025 with 50%+ QoQ growth at ~70% gross margins.
CEO Aidan Gomez, Transformer paper co-author, hired Uber's ex-IPO chief as CFO preparing for 2026 IPO.
Command models run efficiently on limited GPUs; North workspace enables custom secure AI agents.

cohere.com (opens cohere.com)Company profile →

Founded: 20197 yrs
Headquarters: ON
Status: Private
Valuation: $7BprivateSep 2025
Raised: $1.2B3 rounds
Last round: Series E (first tranche) · $500MAug 2025
Works in: AIFrontier models
Coverage: 1 program · 4 updateslatest 1 May 2026checked 11 Oct
Lead investors: Radical VenturesInovia CapitalPSP Investments
Investors: Nvidia (opens nvidia.com)Salesforce (opens salesforce.com)AMD (opens amd.com)
People: Aidan Gomez (opens en.wikipedia.org)CEONick Frosst (opens linkedin.com)Co-founderIvan Zhang (opens linkedin.com)Co-founder

Spec

Command models + North

2025 ARR

$240M^R (reported)

QoQ growth rate

50%+^R (reported)

Gross margin

~70%^E (estimated)

Command A+ parameters

218B (sparse MoE)^R (reported)

Command R+ context window

128,000 tokens^R (reported)