Skip to main content

Moark Provider Guide

Moark is a Tier-2 catalog provider: OpenAI-wire-compatible with no known behavioural quirks, so its entire integration is one JSON file (src/lib/providers/catalog/moark.json) rather than hand-written code. That file is the source of truth for everything on this page.

Verification status: docs- and roster-verified, not yet live-verified. This entry was onboarded credential-free — no API key was created or used. Everything below comes from Moark's public docs (https://moark.com/docs/...) and an unauthenticated GET https://moark.com/v1/models (2026-09-28, HTTP 200, 222 models). No chat/completions request was ever sent, so capabilities, error shapes and per-model context/output limits are not wire-proven. Before relying on this provider in production, run the live-verification steps in Tier-2 onboarding with a funded key.

Key Facts​

  • Provider id: moark
  • Protocol: OpenAI-compatible (/chat/completions) — Moark's docs state "模力方舟的 Serverless API 兼容开发者喜爱且社区流行的 OpenAI 风格 API" (docs/products/apis/texts/text-generation, retrieved 2026-09-28)
  • Base URL: https://api.moark.com/v1
  • Default model: Qwen3-8B
  • Models in catalog: 6 (of 222 live on the full roster — the rest are vision/image/video/audio/embedding/reranking models outside this entry's chat scope)
  • Streaming: documented (stream=True via the OpenAI Python client)
  • Tool calling: documented, model-dependent — a dedicated Function Calling doc page shows the standard OpenAI tools / tool_choice / tool_calls shape, but only specific models are named as supporting it
  • Structured output: not claimed here. Moark documents guided_json / guided_choice (vLLM-style guided-decoding parameters), not the OpenAI response_format field NeuroLink's generic client sends — no response_format wire example exists anywhere in the docs
  • Embeddings: not supported by this entry (embeddings is a separate Moark product/model family, out of scope for the chat catalog)
  • Billing: no free tier documented for API access — prepaid resource packages only (models can be tried free in the web console, but that is not the same as a keyless API tier)
  • Key format: none declared

Quick Start​

1. Get an API key​

  1. Visit: https://moark.com and sign in — no separate signup page is documented, only workbench login
  2. Go to 工作台 → 设置 → 访问令牌 (Workbench → Settings → Access Tokens) at https://moark.com/dashboard/tokens and create a token
  3. Fund a prepaid resource package before making live calls (e.g. the all-model package, https://moark.com/serverless-api/order?package=1910) — billing is prepaid, charged per-call or per-token (https://moark.com/docs/products/apis, retrieved 2026-09-28); models can be tried free in the web console, but no formal free-tier API policy is documented, and package usage is non-refundable (https://moark.com/docs/billing/purchase, retrieved 2026-09-28)
  4. Set MOARK_API_KEY in your .env file

2. Configure​

export MOARK_API_KEY=your-api-key
export MOARK_MODEL=Qwen3-8B            # optional — overrides the default model
export MOARK_BASE_URL=https://api.moark.com/v1   # optional — self-hosted or proxy

3. Use it​

import { NeuroLink } from "@juspay/neurolink";

const neurolink = new NeuroLink();

const result = await neurolink.generate({
  input: { text: "Explain context windows in one paragraph." },
  provider: "moark",
  model: "Qwen3-8B",
});

console.log(result.content);
# CLI
npx @juspay/neurolink generate "Hello" --provider moark

Per-request credentials work as they do for every provider:

await neurolink.generate({
  input: { text: "Hello" },
  provider: "moark",
  credentials: { moark: { apiKey: process.env.MOARK_API_KEY } },
});

Models​

ModelVisionNotes
Qwen3-8B ⭐noRecommended — Moark's own docs name this (and Qwen3-4B) as its suggested free-testing/validation model; the only documented recommendation that also matches a currently-live roster id. Not capability-probed.
Qwen3-4BnoNamed alongside Qwen3-8B as a free testing/validation model; smaller sibling. Roster-listed only; fallback.
ERNIE-4.5-TurbonoBaidu ERNIE 4.5 Turbo — Moark's docs suggest this for multilingual use cases. Roster-listed only; fallback.
GLM-5.3noZhipu GLM-5.3 — on the live roster; not named specifically in Moark's docs. Roster-listed only; fallback.
kimi-k3noMoonshot AI Kimi K3 — closest live match to the docs' "long text" pick (kimi-k2-instruct, no longer on the roster). Roster-listed only; fallback.
DeepSeek-V4-Pro-0813noDeepSeek V4 Pro — closest live match to the docs' "specialized reasoning" pick (DeepSeek-R1, no longer on the roster). Roster-listed only; fallback.

No context window, max output tokens or pricing are shown per model. Moark's docs give only a generic "32K、128K 等" illustration — not tied to any specific model — and the unauthenticated GET /v1/models response returns only {id, object, created, owned_by}, with no context_length or pricing fields. The catalog's defaultContextWindow (32,000) and defaultMaxOutputTokens (4,096) are therefore not vendor-confirmed per-model figures: the context number reflects the vendor's own generic example text, and the output-token number is an unverified conservative placeholder. Both need a live probe (or an authenticated roster call that returns real per-model limits) before they can be trusted.

Fallback order when the default is unavailable: Qwen3-4B → ERNIE-4.5-Turbo → GLM-5.3 → kimi-k3 → DeepSeek-V4-Pro-0813.

The other 216 models on Moark's roster (vision, image, video, audio, embedding, reranking, moderation, TTS/ASR, etc.) are intentionally left out of this catalog entry, which covers only the OpenAI-compatible chat surface.

Verification status​

Tier-2 onboarding requires evidence before a provider is accepted, and pnpm run verify:provider-onboarding gates it in CI. This is what the catalog currently records for Moark — deliberately thinner than a live-verified entry, since no key was created:

ProbeResult
Rosterunauthenticated GET https://moark.com/v1/models, HTTP 200, 222 models, 2026-09-28
Docs cross-checkdocs/openapi/v1, docs/products/apis, docs/products/apis/texts/text-generation, docs/products/apis/texts/function-calling, docs/billing/purchase — all fetched 2026-09-28, confirm OpenAI-compatible wire shape, Bearer auth, and prepaid billing
Auth rejectionnot probed — would require a POST request, out of scope for credential-free onboarding
Live capability sweepnot run — no key was created or used

Because no authenticated or POST request was made, capabilities.tools is recorded as "model-dependent" (the docs name specific Function-Calling models rather than claiming universal support), structuredOutput and structuredOutputWithTools are false (no response_format wire example exists, only the vendor's own guided_json/guided_choice extensions), and errorRules is empty (no documented error-body shape was found).

Troubleshooting​

SymptomCauseFix
Requests fail with an auth-shaped errorMOARK_API_KEY unset, wrong, or the account has no funded packageCheck the token at https://moark.com/dashboard/tokens and confirm a prepaid package is active
Model not foundThe roster changed since 2026-09-28Pick a current id from the unauthenticated GET https://moark.com/v1/models roster
Structured output not returned as expectedThis entry doesn't send response_format — Moark only documents guided_json/guided_choiceNot currently wired through the generic OpenAI-compat client; treat structuredOutput as unsupported until a live probe proves otherwise
Tool calls silently ignoredFunction calling is documented as model-dependent, not universalConfirm the chosen model is one of Moark's documented Function-Calling models before relying on tools

See also​