Cohere vs Google DeepMind

Introducing North 2: AI without compromises

Cohere has announced North 2, its latest platform update with 15+ new features, positioned as enterprise AI without compromises.

North 2: Enterprise AI without compromises

Cohere's biggest upgrade yet — unifying enterprise-grade security, performance, and flexibility in one platform.

Aleph Alpha's Kolibri Lags Behind Open-Weight Leaders

Aleph Alpha has released Kolibri, a new open-source German-English AI model under the Apache-2.0 license, designed for mission-critical applications in public administration and industry. Kolibri features a Mixture-of-Experts architecture with 78 billion total parameters and about 3.46 billion active per token. Trained on 20 trillion tokens, with 23% in German and a German-optimized tokenizer, it supports context lengths up to one million tokens. The model excels in agentic workflows, RAG, and native tool calling, while emphasizing data sovereignty and compliance, including a blocklist of 4.5 million URLs and GDPR/EU AI Act adherence. Independent benchmarks, however, place it behind at least 20 other open-weight models. Aleph Alpha is merging with Cohere to form a transatlantic sovereign AI company, subject to regulatory approval. In a broader economic debate, Reid Hoffman argues AI infrastructure expansion is the 'only reason' the US avoids recession, though economists dispute this.

• Aleph Alpha released Kolibri under Apache-2.0, with 78B total parameters and ~3.46B active per token.
• Kolibri trained on 20T tokens, with 23% German data and a German-optimized tokenizer, supporting 1M context.
• Independent benchmarks rank Kolibri behind at least 20 other open-weight models.

Google DeepMind alum launches London AI reasoning company Metis

Former Google DeepMind research lead and AlphaGo co-creator Thore Graepel has launched Metis Reasoning, a London-based AI company pursuing a compute-efficient alternative to frontier AI models. The startup focuses on a different approach to AI reasoning, aiming for verifiable, robust, and safe outputs, distinct from large language models. The founding team includes Piers Ingram, Michael Warren, and David Berman. Metis has reportedly raised tens of millions of dollars, though Ingram declined to confirm. The company believes its approach lends itself to industries like robotics and engineering. Metis joins other new London AI ventures, including Ineffable Intelligence and Recursive Superintelligence.

• Thore Graepel, co-creator of AlphaGo, left Google DeepMind in summer 2026 to launch Metis Reasoning.
• Metis is based in London and has a 12-strong team of researchers, mathematicians, robotics experts, and tech entrepreneurs.
• The founding team includes Thore Graepel, Piers Ingram, Michael Warren, and David Berman.

Elorian launches Glimpse AI model for visual reasoning

Palo Alto-based AI startup Elorian, which previously raised a $55 million seed round at a $300 million valuation, has released its first AI model, Glimpse, now available as a research preview with API access. Glimpse is designed for visual reasoning, utilizing a 'visual scratch space' that allows the model to crop into images, extract video frames, or run code to analyze visual content during reasoning. According to Elorian's evaluations, Glimpse outperforms GPT-6 Astra on scientific chart reasoning and short-form video captioning. The company, led by former Google DeepMind researcher Andrew Dai, targets companies with substantial video assets, including robotics, scientific labs, media, and creative sectors. Pricing is set at $0.60 per million input tokens and $3 per million output tokens. Elorian plans to build what it calls 'the world's best visual reasoning model' next year.

• Elorian, a Palo Alto AI startup, launched its first AI model, Glimpse, as a research preview.
• The model uses a 'visual scratch space' to crop into images and extract video frames during reasoning.
• According to Elorian, Glimpse outperforms GPT-6 Astra on scientific chart reasoning and short-form video captioning.

Nvidia, Google DeepMind, Mistral Support EU AI Research Initiative

The European Commission has launched the RAISE initiative (Resource for AI Science in Europe) to boost AI research in Europe. The program provides research institutions with access to computing power, AI models, data, and technical expertise. Nvidia, Google DeepMind, Mistral, Elsevier, Sanofi, Frontiers, and the Novo Nordisk Foundation have committed support. A notable project is the collaboration between Mistral AI and CERN to develop a large AI foundation model for physics. The initiative also includes three digital platforms: RAISE Hub, RAISE Forum, and RAISE Observatory. The Commission also updated guidelines for responsible AI use in research, now covering agentic AI. This is part of the European AI science strategy announced in October 2025.

• The European Commission launched the RAISE initiative to support AI research in Europe.
• Initial support commitments come from Nvidia, Google DeepMind, Mistral, Elsevier, Sanofi, Frontiers, and the Novo Nordisk Foundation.
• Mistral AI and CERN are collaborating on a large AI foundation model for physics.