Open source business models in 2026: a field guide

Quick answer. Open source monetization in 2026 has consolidated around five viable models: (1) open-core (free core, paid enterprise features), (2) SaaS (managed cloud version of an open-source tool), (3) dual-license (AGPL for community, commercial license for closed-source users), (4) support contracts (SUSE, Red Hat, MongoDB-style), and (5) foundations (CNCF, Linux Foundation, Apache). The new wave in 2026: AI-specific models (per-token API, fine-tuning services, hosted model serving). The 90% answer for new projects: open-core with a SaaS cloud version.

The 2026 landscape

There are roughly 50,000 active open-source projects on the GitPop radar in 2026 H2. Of those with a viable business model:

• ~40% are hobby projects with no business model
• ~25% are funded by foundations (CNCF, Apache, Linux Foundation, etc.)
• ~20% are open-core with a SaaS cloud version
• ~10% are SaaS-only (the open-source version is a marketing funnel)
• ~5% are dual-license (AGPL + commercial)

The numbers are estimates from the GitPop radar data, cross-referenced with Crunchbase and the project README files. The categories are not exclusive — a project can be open-core, dual-license, and foundation-funded at the same time.

The five viable models

1. Open-core

The default for most successful open-source projects in 2026. The open-source core is feature-complete for individual developers and small teams. The paid enterprise tier adds features that large companies need: SSO, audit logs, RBAC, dedicated support, compliance certifications.

Examples: GitLab, Sentry, GitHub, Supabase, PostHog, Plane, Cal.com, AppFlowy.

When to pick: when the enterprise features (SSO, audit logs, compliance) are a meaningful additional value. If the core is already enterprise-ready and there's no clear "enterprise-plus" tier, open-core is the wrong model.

2. SaaS (managed cloud)

The second-most-popular model. The open-source version is self-hostable; the SaaS version is a managed cloud. The SaaS version is the business; the open-source version is the marketing funnel.

Examples: Supabase (managed Postgres), Vercel (managed Next.js), PlanetScale (managed MySQL), Neon (managed Postgres), Inngest (managed background jobs), Trigger.dev (managed background jobs), Meilisearch Cloud, Typesense Cloud.

When to pick: when the operational complexity is high enough that most users will pay for a managed version. The economics: 5-10% of self-hosters convert to paid cloud. The conversion is high enough to make the SaaS business viable when the underlying software is complex.

3. Dual-license

The original open-source business model. The software is available under AGPL (or a similar copyleft license) for the open-source community, and under a commercial license for companies that don't want to open-source their modifications.

Examples: MongoDB (now SSPL), MariaDB (BSL), CockroachDB (BSL), Elastic (SSPL), Grafana (AGPL), Mattermost, Nextcloud, Plausible.

When to pick: when the software is infrastructure (database, search, cache) and SaaS hosting is a real threat to the business model. The license change from AGPL to SSPL/BSL is increasingly common as companies move to defend their SaaS revenue.

The risk: the license change alienates the open-source community. MongoDB's SSPL transition in 2018 was controversial; Fork (the community fork) still exists. Elastic's SSPL transition in 2021 had a similar backlash. The model works financially, but the social cost is real.

4. Support contracts

The original enterprise open-source model. The software is free; the company charges for support, training, and consulting.

Examples: Red Hat (RHEL), SUSE (SUSE Linux), Canonical (Ubuntu), Cloudera (Hadoop), MongoDB (Atlas support).

When to pick: when the software is operationally complex enough that enterprises will pay for support. The model is harder to scale than SaaS or open-core — the marginal cost of a support contract is high (engineer time), so the model only works for high-value enterprise customers.

5. Foundations

The software is donated to a foundation (CNCF, Apache, Linux Foundation, etc.) and developed by a community. The foundation provides legal cover, governance, and marketing. The business model is usually ecosystem — companies contribute to the foundation in exchange for hiring, influence, and ecosystem access.

Examples: Kubernetes (CNCF), Apache Kafka (Apache), TensorFlow (LF AI), Postgres (PgFoundation), Linux (Linux Foundation).

When to pick: when the project is large enough to attract a community of contributors and corporate sponsors. The model is uncommon for small projects; it becomes attractive when the project hits ~100 contributors and ~$1M+ in ecosystem revenue.

The new wave: AI-specific models

The 2026 wave of open-source business models is specifically for AI. The leading patterns:

• Per-token API — the open-source model is free; the hosted API is paid per token. Used by Mistral, DeepSeek, and most of the open-weight model providers.
• Fine-tuning services — the open-source model is free; the company provides fine-tuning, evaluation, and hosting. Used by Hugging Face, Replicate, and Modal.
• Hosted model serving — the open-source serving engine (vLLM, llama.cpp) is free; the managed cloud serving is paid. Used by Anyscale (vLLM), Together AI, Fireworks AI.
• AI-specific SaaS — the open-source LLM tooling is free; the SaaS wrapper (with extra features) is paid. Used by the entire AI app ecosystem (Perplexity, Cursor, etc.).

The AI models are different from the traditional open-source models in one important way: the cost of running the software is high (GPU time is expensive). This shifts the economics toward SaaS and away from self-hosting. The 5-10% self-host-to-SaaS conversion of the traditional model is closer to 1-3% for AI because the cost of self-hosting is high.

What is next

The 2026 H2 outlook:

1. Open-core will remain the default. The economics work, the community is happy, the enterprise customers are willing to pay. Most new projects will follow this pattern.
2. Dual-license will continue to migrate to BSL/SSPL. The pressure from SaaS competition is real. Expect more AGPL projects to switch to BSL or SSPL in 2026-2027.
3. AI-specific models will mature. The per-token API model is working. The fine-tuning service model is working. The hosted model serving model is working. The composability of these models is the next thing to figure out.
4. Foundations will continue to be the home for infrastructure. Kubernetes, Postgres, Kafka, Spark — the infrastructure tier is increasingly foundation-managed. The trend will continue.

For the broader data, see State of GitPop H2 2026. For the framework decision, see The end of monolith frameworks. For the AI toolchain, see Building your AI dev stack in 2026.