The Lab
Founded in 2023 as an independent research collective rather than a venture-backed lab, Nous Research occupies a unique position in the AI landscape. The structure is intentionally small, the team is distributed, the output is craftsmanship-focused — meticulously fine-tuned models released as open weights, designed for the agentic-builder community that values control over scale.
The 2026 flagship is Hermes 4 — an open-weights agent model with native persistent memory, designed to compound intelligence as it accumulates skills and conversations. Hermes Vision 2 added multimodal screen-reading for vision agents. The cost story is the differentiator: both models are free to download, free to run on any hardware that can host them, with the indie-builder community running them on $9/month Hostinger VPS instances.
The catch is exactly the strength. Nous is not a venture-backed frontier lab and isn't trying to be. There is no Hermes 5 racing to match GPT-5.5 on every benchmark. The models won't be the absolute best at any single category. What they will be is craftable, ownable, and ecosystem-friendly — characteristics that matter to the indie builder economy but not necessarily to enterprise procurement.
Recent Coverage
Models
Open-weights agent model with native memory architecture. Free to download and run, popular with the $5/month-VPS builder ecosystem. Compounding intelligence as it accumulates skills.
Best for
Open-weights multimodal variant. Stronger than most open vision models on chart and screenshot understanding. Common pairing with Hermes for agent screen-reading workflows.
Best for
When to Pick Nous Research · When to Pick Someone Else
Each lab in the atlas comes with its own positioning, model line, and use cases. The point of organising the AI landscape by lab is that the answer to "which model should I use" almost always starts with "which lab is closest to what I'm trying to do."
