A Canadian model developer. Rather than building consumer chatbots, it builds models that enterprises and governments run on their own servers. Its flagship models are the Command series of language models. In August 2026, the company expanded its lightweight model lineup by unveiling North Micro Vision, a 2.4-billion-parameter vision model. That same month, Cohere's CEO shared his views on how models should be opened up, laying out three conditions he believes an open-source model must meet.
Current rank (1M)
23
Rank over the last 30 days
2026.08.20 – 2026.09.18 · High 22 · Low 47 · Now 23
Cohere CEO Aidan Gomez said live on Bloomberg TV on September 14 that the existential-risk debate around AI has veered too far into science fiction. He cited the Hugging Face hacking incident as an example of over-extrapolation and called it a red flag for the largest companies to set the rules among themselves.
Parse 5, the document-parsing vision language model Cohere released on August 27, scores 79.2 on ParseBench, below GPT-5.5 and Opus 4.8. The company published that table itself and led instead with throughput and cost: 2,160 pages a minute on eight H100s.
In a statement posted to X on September 14, the OpenAI CEO welcomed a federal safety framework but said the company will not wait for an antitrust exemption or legislation. He also disclosed a new procedure: writing a safety case before any training run expected to sharply raise capabilities.
Aidan Gomez, CEO of Cohere, published an essay on September 14 calling Anthropic's proposal to pace the frontier a cartel. Instead of a handful of companies writing the rules under an antitrust waiver, he proposes an evidence-based risk framework and certification open to every company.
Even after silencing unspoken numbers, some models still transcribed them correctly 30-40% of the time
Bloomberg interviewed the head of LG AI Research's Superintelligence Lab on-site at AI Summit Seoul. Why did the leader of a team that had just passed the government's second-round evaluation for a homegrown model choose the word "sovereign," of all things?
Ultra-compact VLM specialized in document and chart understanding, released free on Hugging Face under Apache 2.0
"DLR-Lock" research lets users run open-weight models as-is while blocking modification
Aidan Gomez explains why the company is expanding its open-source lineup and how it's designing for "sovereign" deployment
AWS unveiled a native search feature that manages operational data and embeddings together without requiring a separate vector database
That's the last story.
