Basecamp Research
Biological foundation models · Health
How they use Nebius AI Cloud
Basecamp Research trained EDEN, its biological foundation model, on Nebius AI Cloud. EDEN learns from genetic sequences to support medicine research. Nebius’s July 22 post describes a 28-billion-parameter model trained on 9.7 trillion tokens across 1,028 NVIDIA H200 GPUs, completing pre-training in six weeks. It reports 20x faster training and lab-validated progress in gene insertion and treatments for drug-resistant bacteria. Nebius’s shareholder letter says training ran entirely on Nebius and identifies the speed comparison as Basecamp’s previous internal environments. Nebius also says the EDEN family is producing new scaling laws for biology.
Basecamp’s customer presentation explains how it used Nebius storage for large biological datasets. An official June 2026 customer video also describes InfiniBand networking and Managed Kubernetes supporting training. Nebius’s July 9 post describes petabytes of sequence data covering 10 billion new-to-science genes across 1 million new species. Nebius says it supplied the large-scale compute within weeks. High-performance, low-latency storage fed a 1,024-GPU cluster that stayed stable for six weeks.
Nebius named Basecamp among the first partners for its UK deployment. Co-founder Oliver Vince said Nebius infrastructure provided the scale and performance Basecamp needed and that UK availability gave it greater control over its data.
Sources
5 sourcesNebius EDEN training update
9 Jul 2026 · LinkedIn · Nebius EDEN training update · Web page
Nebius says Basecamp Research trained EDEN on its infrastructure using petabytes of biological sequence data. It supplied compute within weeks and low-latency storage supporting a stable six-week run across 1,024 GPUs. Nebius describes itself as a training home for the EDEN model family and says the models are producing new scaling laws for biology and supporting antibiotics and gene-editing therapeutics.
Open original sourceNebius Making Medicine Programmable: Basecamp Research Trains AI Model Eden 20x Faster on Nebius
23 Jun 2026 · Nebius · Video
1:54–3:59: Basecamp Research describes partnering with Nebius for GPU training compute and training the EDEN foundation-model family. The presentation describes large-scale storage and up to 1,024 GPUs for the historical work, not current capacity.
Open original source at 1:54Nebius Describes training EDEN on a Nebius H200 cluster
22 Jul 2026 · Nebius · Web page
Nebius says Basecamp Research trained EDEN on Nebius AI Cloud using 9.7 trillion tokens across 1,028 NVIDIA H200 GPUs. The 28-billion-parameter model completed pre-training in six weeks. Nebius reports 20x faster training and lab-validated progress in gene insertion and treatments for drug-resistant bacteria.
Open original sourceNebius UK AI Cloud launch: Basecamp Research
6 Nov 2025 · Nebius · Web page
Nebius names Basecamp Research among the first partners for its UK deployment. Basecamp co-founder Oliver Vince says Nebius infrastructure delivers the scale and performance the company needs and that UK availability gives it greater control over its data.
Open original sourceCompany context
basecamp-research.comBasecamp Research develops biological AI models and therapeutic candidates for pharmaceutical partners and researchers. Its EDEN foundation models use proprietary genomic data collected through international biodiversity partnerships to design DNA sequences, proteins, enzymes and peptides. The company offers candidate development, joint research programmes and access to its design platform. Researchers can also access antibiotic design and antigen immunogenicity prediction through Claude. Its internal therapeutic pipeline covers gene insertion, gene editing and peptide design, including preclinical work on cell therapies that modify cells inside the body.
Basecamp ResearchRecent updates
1 updateBasecamp Research raises $140 million Series C
Basecamp Research announced $140 million in Series C funding to advance AI-designed therapeutics. Its distributor corrected part of the release on September 24.
Basecamp Research corrected release