Quick answer: Onyx AI is free to self-host under its MIT-licensed Community Edition, and Onyx sells two paid plans. Business costs $20 per user per month billed annually or $25 billed monthly, and Enterprise is custom-quoted. Zoom acquired Onyx in September 2026, and the Onyx pricing page still lists the same rates. But the license price is only part of the bill. Running Onyx yourself means paying for servers, for local GPUs or LLM API usage, and for engineers who patch, monitor, and upgrade the deployment. By Onyx’s estimate, local model hardware runs $3,000 to $5,000 for a small team and $120,000 or more for a large one. A 100-person team on cloud LLM APIs spends roughly $1,000 to $5,000 a month. That gap between license cost and total cost of ownership, or TCO, is the real number to plan around. It is the gap a fully managed platform like GoSearch is built to close.
Pricing and product details sourced from the Onyx pricing page, Onyx documentation, and the Onyx AWS Marketplace listing. Last updated October 2026.
Onyx AI’s Published Pricing
How much does Onyx AI cost?
Onyx offers a free Community Edition and two paid plans. The Community Edition is free under an MIT license and covers the core RAG, AI chat, agent, and action features, according to the Onyx GitHub repository. You host it yourself. Business costs $20 per user per month with annual billing or $25 per user per month billed monthly, per the Onyx pricing page. Onyx says Business has no seat minimum and no mandatory support fee, and Onyx Cloud offers a two-week free trial with no credit card. Enterprise is quoted individually. It adds OIDC and SAML single sign-on, on-premise and region-specific deployments, white-labeling, invoice billing, volume discounts, and dedicated SLA-backed support.
Some aggregators list older or incorrect figures, such as $16 per seat per month or $20 per user per year. The current published rate as of October 2026 is $20 annual or $25 monthly.
What’s the difference between Onyx Cloud and the Community Edition?
Onyx Cloud is the managed hosting option. The Community Edition is software you deploy on your own Docker or Kubernetes environment and run yourself. Both include the core features, such as connectors, AI chat, agents, search, and deep research, according to Onyx documentation. The paid plans add the controls larger organizations typically need. Onyx documentation lists document permission inheritance, role-based access control, query history and usage dashboards, and encryption of secrets as Business features, so the free edition does not include them. Enterprise adds SCIM group sync, full white-labeling, custom roles, configurable usage limits, custom deployments, region-specific data processing, and enterprise SLAs.
Teams that self-host and want those features pay for a license, which Onyx issues to the instance after they subscribe. Onyx documentation lists self-hosting as an optional Business feature, while the Onyx pricing page lists on-premise deployments under Enterprise. Confirm which plan covers your deployment before you budget.
What does Onyx Enterprise Edition actually cost?
Onyx doesn’t publish an Enterprise price. It is quote-based. The AWS Marketplace listing for Enterprise Edition shows how the contract is structured. Onyx charges on two dimensions: the number of users with access and the number of Onyx instances you run. User count usually drives most of the cost. Instance count adds cost when you run extra separate deployments, such as regional or isolated environments. The listing also notes that additional AWS infrastructure costs may apply. An organization that runs three regional instances should budget for all three on top of its user count. That detail is easy to miss when you budget from the $20 headline rate.
Onyx AI Pricing After the Zoom Acquisition
When did Zoom acquire Onyx?
Onyx co-founders Chris Weaver and Yuhong Sun announced on September 23, 2026 that their company is joining Zoom. Zoom did not disclose the terms. The 18-person team joins Zoom AI Services, Zoom’s developer platform, according to coverage of the announcement. Zoom said the team will help it serve customers that need on-premises deployment and strict data sovereignty, per Dealroom. Weaver said Onyx stays “open source, model agnostic, app agnostic, and self-hostable.” Sun said customers will keep receiving the same product from the same team.
Did Onyx pricing change after the Zoom acquisition?
No pricing changes have been announced. The Onyx pricing page still lists Business at $20 per user per month with annual billing and Enterprise as custom. Zoom has not published a priced press release, according to independent analysis. It is also not yet clear whether Onyx stays a separate product line or becomes part of a larger Zoom bundle. Buyers who sign multi-year agreements can ask Onyx or Zoom to confirm in writing how pricing and packaging will work through renewal.
What does the Zoom acquisition mean for self-hosting Onyx?
Nothing announced so far changes the self-hosting model. The Community Edition stays free and MIT licensed, and Onyx says the product remains self-hostable. The cost drivers covered below stay the same: servers, LLM usage, and ongoing operations. A larger parent company does not remove the work of running the deployment. One analysis of the deal notes that an announcement can make promises but cannot guarantee them. Teams whose requirements depend on self-hosted or air-gapped deployment can ask for written confirmation that those options stay available through their agreement.
The Real Cost of “Free”: Self-Hosting Onyx’s Community Edition
Is Onyx AI’s Community Edition really free?
The software license costs nothing. Running it in production does. Onyx documentation says self-hosting “requires more operational overhead compared to the Cloud offering.” Onyx generally recommends Cloud unless a team specifically needs to self-host. The same page says teams can self-host the free edition to reduce costs if they don’t need Enterprise Edition capabilities, but they give up certain advanced security features.
The free edition also leaves out document permission inheritance and role-based access control. Those two features decide whether an AI search tool shows each employee only what they can already open in the source system. Teams that need them pay for a Business or Enterprise license, even when they host Onyx themselves.
What does a self-hosted Onyx deployment involve?
Onyx supports two deployment modes, according to the Onyx GitHub repository. Onyx Lite is a lightweight chat interface that runs in under 1 GB of memory but cannot index documents. Standard Onyx is the full platform, and it is the mode that delivers enterprise search. A standard deployment runs a vector and keyword index, background workers that sync data from connectors, model inference servers for indexing and retrieval, a Redis cache, and MinIO blob storage. Teams deploy it with Docker Compose, Kubernetes, Helm, or Terraform.
Onyx documentation says upgrades usually need only a redeployment command. The effort sits in monitoring, scaling, backups, and security patching across each of those components.
How much does self-hosting Onyx cost in infrastructure?
The answer depends on where the LLM runs. Teams that call cloud LLM APIs skip GPU purchases and pay per token instead. Onyx estimates that a 100-person team averaging 50 queries a day spends $1,000 to $5,000 per month on API costs, depending on the provider and model. Teams that run models locally buy hardware. Onyx publishes these estimates:
- Small team, 5 to 20 users: one high-end consumer GPU or a Mac Studio, roughly $3,000 to $5,000 one-time
- Mid-size team, 20 to 200 users: one H100 or two high-end consumer GPUs, roughly $25,000 to $35,000 one-time
- Large team, 200 to 1,000 or more users: multi-GPU clusters, $120,000 to $500,000 one-time, or $8,000 to $20,000 per month on cloud GPU instances
These estimates cover model-serving hardware. They do not include the servers that run Onyx itself, backups, or monitoring. Air-gapped deployments must run local models, because Onyx documentation says an offline setup needs your own LLM.
How much engineering time does self-hosting Onyx require?
Initial setup is short by Onyx’s account. The Onyx self-hosting guide says a Docker Compose deployment reaches a first query in under an hour. It estimates 2 to 4 hours for a mid-market deployment and 1 to 2 days for an enterprise deployment, followed by 1 to 2 weeks to roll out connectors and onboard users.
Ongoing work is where the cost builds. Someone has to monitor the deployment, schedule upgrades, apply security patches, scale the index as connectors and users grow, and tune retrieval when answers miss. The guide notes that tuning retrieval quality and evaluating model outputs can call for ML expertise. It does not give an estimate for ongoing staffing, so budget that from your own DevOps and engineering rates.
The guide reserves its 2 to 4 month estimate for a fully DIY stack built from separate inference, chat, retrieval, and authentication tools. That is the build Onyx is designed to replace, and teams that deploy Onyx skip most of it.
When does self-hosting models break even against cloud LLM APIs?
For a 100-person team, Onyx’s guide cites a 6-to-12-month break-even for self-hosted GPU inference against cloud API costs, assuming moderate usage. That math covers model inference only. It excludes the staff time and servers needed to run the platform around the model.
Onyx AI Pricing vs. GoSearch
Is GoSearch a lower-TCO alternative to self-hosting Onyx?
For teams that want grounded AI search across company data without running the infrastructure, GoSearch removes the server, GPU, and operations line items from the calculation. There are no workers to monitor, no indexes to size, and no upgrades to schedule. GoSearch runs on a hybrid, real-time federated architecture, so queries pull live data from connected systems instead of requiring a team to stand up, secure, and maintain a separate vector database and indexing pipeline.
GoSearch prices per seat with published, flat pricing rather than a quote-only Enterprise tier. The number on the pricing page is closer to the number on the invoice. For a more in-depth breakdown, see a comparison of GoSearch vs. Onyx.
When does Onyx make more sense than GoSearch?
Self-hosting buys maximum infrastructure control, which some organizations genuinely need. Onyx fits teams that require a fully air-gapped deployment with local models, teams with DevOps capacity that want source-level control over the platform, and teams that want a free open-source starting point. Zoom’s acquisition puts a larger vendor behind that model, and Zoom has positioned Onyx for customers with on-premises and data sovereignty requirements.
Teams choosing based on total cost, time-to-value, and not wanting to own the operational overhead of a self-hosted AI stack are the ones GoSearch is built for. Teams that need more control over where their data lives can also use GoSearch’s BYOC deployment option.
How does Onyx AI’s pricing compare to Glean?
Onyx also markets against Glean, another indexed enterprise search platform. In the Onyx Glean cost calculator, a 250-seat deployment on Onyx Business runs $60,000 a year, or $180,000 over three years. The calculator estimates Glean at $182,000 in year one, rising to $220,000 in year three, for a three-year total of $601,000. It builds those figures from publicly reported Glean rates of $45 to $65 per user per month, a support fee of about 10 percent, contract minimums of $50,000 to $60,000, and annual renewal increases of 10 percent.
At 500 seats, Onyx lists $120,000 a year against roughly $300,000 to $430,000 for Glean. Onyx built these estimates from public reports rather than confirmed Glean quotes, so treat them as directional. The Onyx figures reflect the Business license price. Teams that self-host take on the infrastructure and operations costs covered above, which the calculator leaves out.
Key Takeaways
- Onyx’s Community Edition is free and MIT-licensed, but free refers to the software license, not the total cost of ownership.
- Onyx sells two paid plans. Business costs $20 per user per month billed annually or $25 billed monthly, and Enterprise is custom-quoted and billed on both user count and instance count.
- Zoom acquired Onyx in September 2026. Onyx pricing is unchanged, Onyx says the product stays open source and self-hostable, and Zoom has not published packaging details.
- The free edition does not include document permission inheritance or role-based access control. Self-hosting teams that need them pay for a Business or Enterprise license.
- Self-hosting costs servers plus LLM usage: $3,000 to $5,000 in hardware for a small team up to $120,000 or more for a large one when running local models, or roughly $1,000 to $5,000 a month in API costs for a 100-person team, per Onyx’s estimates.
- Onyx puts initial deployment at hours to days. Its 2 to 4 month estimate applies to a DIY stack, so the cost of self-hosting Onyx sits in running it, not installing it.
- Self-hosted inference typically breaks even against cloud APIs in 6 to 12 months, but that timeline covers model costs only, not platform operations.
- A fully managed platform like GoSearch removes the infrastructure and operations line items from the TCO calculation, trading self-hosting control for published, flat per-seat pricing and no servers to run.
- Onyx has no user reviews on Capterra or G2, so independent feedback on satisfaction and costs at scale is limited.
