Skip to content
Heimdallm
Marketplace/Documentation
Documentation · Guides

Requesting a model

When a model you want is not in the catalogue, ask for it on the model requests page. A request is a Hugging Face link and nothing else: the marketplace reads the model's name, size, context window, what it reads and makes, and its license from the model card, so there is no form to fill in.

Asking for a model

Paste the model's page, huggingface.co/owner/model, or its id, owner/model, into the request box. The card under the box shows what the model card states. Then:

  • Not listed and not asked for yet: requesting it opens the request.
  • Already requested: requesting it adds your vote to the open request.
  • Already listed: the box links the model page instead.

A quantized upload counts toward the model it was made from. A request for someone/Model-32B-GGUF whose card names owner/Model-32B as its base is a request for owner/Model-32B, because providers choose the quantization they serve. A fine-tune, a merge or an adapter is a different model and is requested on its own.

Any signed-in account can open requests, up to ten a day. Posts in the requests forum on Discord that link a Hugging Face model open requests too.

Voting

Votes rank the open requests, and providers read the ranking to decide what to list next. A vote counts once per account and only from an account holding credit: it is a customer saying what they would spend on. Requesting a model from an account holding credit casts that account's vote on it. Votes are anonymous; the marketplace stores a keyed hash of the account rather than the account itself.

How a request moves

StateWhat it means
OpenWaiting for a provider.
Being listedA provider took it on, or a provider's listing already declares the model.
LiveAn endpoint for the model passed its first health check. The badge links the model page.
DeclinedThe marketplace will not list it, with the reason shown on the badge.

Every step comes from what the marketplace observes, not from anyone marking it done. The time shown on a live request runs from when it was asked to when its first endpoint passed health, and the figures at the top of the page (median time to live, share of decided requests that went live) are computed from those same moments.

Next

DiscoverFind your next modelModel libraryAll models, versions, prices and capabilitiesBenchmarksMeasured quality, cost and time by serving configurationModel requestsAsk for a model, and vote on what is listed nextShowcaseImages, video and 3D made on the marketplaceProvidersCompare independent inference providersActivityMarketplace traffic and usageNetwork statusAvailability, latency and endpoint healthChatConversations, model settings and filesImage, video & 3DGenerate images, video and 3D assetsScene studioStage a shot in 3D, direct it and render it as a videoCoding playgroundBuild software with an AI agent on a machine of your ownOverview & routingIdentity, workspaces and routing preferencesBalance & depositsBalance, deposits, ledger and reserved fundsAPI keysKeys scoped to workspaces, limits and expirationOrganizationsTeams, members and shared workspacesRequest historyRequests, receipts, sandbox sessions and account historyReferralsInvitations and referral earningsPrivacyData settings, exports and deletionDocumentationQuickstart, SDKs, pricing and guidesAPI referenceInteractive endpoint referenceList your modelsProvider guide: requirements, the fee, payouts and what is measuredProvider payoutsMonthly statements, the marketplace fee and payout walletsProvider dashboardApply as a provider, add sources and track earnings
Pages, models and API toolsESC to close