ElevenLabs Voice Agent Launched on OpenAI Marketplace
ElevenLabs announced that its conversational agent product, ElevenAgents, has entered the enterprise marketplace launched by OpenAI at DevDay: OpenAI enterprise customers will be able to purchase using existing commitment credits and equip the system with voice in over 90 languages; the catalog is now live, and ordering functionality will soon be available, with interest registration currently open.
The company stated that the product combines OpenAI's model intelligence with its own real-time human-like voice stack, covering listening, reasoning, responding, and action, all in one purchase. Ashish Uchil, head of business and partnerships, expressed that intelligence comes from OpenAI, while voice and dialogue stack come from ElevenLabs. The product page and documentation separately list real-time voice, integration, knowledge base, guardrails, and analytics, as well as deployment paths like web components, mobile SDK, and SIP trunking; language models can be swapped or brought in.
Recently, ElevenAgents integrated with Eleven v4 Turbo, promoting a median reasoning latency of about 100 milliseconds, and has pre-configured scenarios such as reception and sales development representatives. The voice library and language count vary across different pages, ranging from over 70 to over 90 languages and thousands to tens of thousands of voice tones, with the external standard based on market posts being over 90 languages. The marketplace is only open to OpenAI enterprise customers, and consumer-level ChatGPT users are not included in this commitment credit range.
The OpenAI marketplace is described as an enterprise tool shelf, with ElevenLabs being one of the launch partners. For companies that have locked their budgets in OpenAI contracts, there is no need to open a separate main vendor account for the voice layer; they can simply add an item to the same commitment.
In terms of market mechanics, buyers are teams with existing OpenAI enterprise contracts but lack deployable voice seats, while sellers are those adding voice tones and multilingual capabilities to existing commitments. Funds flow from OpenAI enterprise credits to ElevenLabs without needing a new procurement round; beneficiaries are voice layer vendors and cloud model providers looking to increase contract utilization, while those under pressure are other voice agents that must sign separate contracts and cannot tap into the same commitment pool. The event driver is the DevDay shelf debut, not a new model ranking.
Commitment credits have become shelf currency: model vendors provide settlement, while voice vendors provide the last mile of oral communication. Whoever gets on this shelf first will be the first to collect the unspent money in enterprise contracts.
Source: Public Information
ABAB AI Insight
ElevenLabs has transitioned from a voice API to hosting conversational agents and now into others' enterprise markets, leveraging its strongest asset—low-latency multilingual oral communication—onto the largest enterprise model bills. The ability to swap in its own models indicates it is not competing with OpenAI for the base but for the "voice output" layer. The DevDay shelf shifts the plugin logic from ChatGPT conversations to enterprise procurement: catalog, commitment credits, and soon-to-be-available ordering represent the Salesforce app store in the model era. The historical path started with creator voiceovers, then customer service and reception agents, and now leveraging platform distribution to avoid building sales for each industry individually.
The capital path is embedded distribution. Enterprises have already signed annual commitments to OpenAI; if the voice agent requires a separate contract, it must go through security and procurement again; post-listing, security reviews and payment account reuse compress the sales cycle to catalog clicks. OpenAI increases commitment consumption, while ElevenLabs gains default voice positions. The motivation is that phone and multilingual customer service remain the dirtiest last mile for agents, which pure text agents cannot cover; whoever controls the oral communication controls whether these work orders truly leave the keyboard.
Analogous examples include Twilio becoming the telephony layer in the cloud and Zoom integrating telephony into meeting packages. The industry phase is enterprise agents moving from chat boxes to callable entities. Companies developing their own voice models will keep oral communication within their own stack; independent voice companies must become shelf SKUs or risk missing out on locked commitment pools. Reception and SDR templates are designed to make procurement reasons specific to roles rather than just "buy another API."
Structural judgment indicates a reconstruction of the industry chain. The mechanism centralizes settlement rights to model contracts, transforming capability layers into shelf products: voice, search, and browser operations will all be charged against the same commitment. Reconstruction occurs because enterprises are unwilling to sign a main vendor for every agent capability. Pricing power temporarily rests with the shelf owner—whoever decides on listings and revenue sharing determines the growth rate of oral communication vendors.
ABAB News · Cognitive Laws
Enterprise commitment credits are the new shelf currency
Models handle the intelligence, while oral communication can still be a standalone business
First, get on someone else's bill, then negotiate your own pricing power