Skip to main content

Subconscious Provider Guide

Subconscious is a Tier-2 catalog provider: its integration is one JSON file (src/lib/providers/catalog/subconscious.json) rather than hand-written code. That file is the source of truth for everything on this page.

Verification status: this entry is docs- and roster-verified, not yet live-verified. It is built from Subconscious's public documentation and pricing pages and unauthenticated GET requests to /v1/models and /v1/chat/completions. No API key was created or used, and no POST request was sent. evidence.liveMatrix is null until someone runs the live capability matrix with a real key (see Verification status below).

Key Facts​

  • Provider id: subconscious
  • Protocol: OpenAI-compatible (/chat/completions). The vendor overview says "Our API is compatible with both the OpenAI Completions and Anthropic Messages APIs" (overview); this entry uses the OpenAI format
  • Base URL: https://api.subconscious.dev/v1
  • Default model: subconscious/glm-5.3-marathon
  • Models in catalog: 3 (the roster lists 4)
  • Streaming: supported. The API reference says "Each event is a ChatCompletionChunk object" and describes the stream as ending with data: [DONE] (API reference)
  • Tool calling: declared false; not probed with a key. The pricing page's plan card says "Drop-in for Claude Code, Codex, OpenCode, or any OpenAI-compatible harness" (pricing)
  • Structured output: declared false. The API reference says of response_format: "Currently unsupported. The request may accept this field, but it is ignored and does not constrain the model output to JSON or a schema."
  • Structured output + tools together: declared false; not probed with a key
  • Thinking: declared false. The API reference lists chat_template_kwargs with enable_thinking as a "Subconscious extension" (API reference)
  • Embeddings: declared false
  • Billing: no-free-tier is the catalog schema's required field value here, not a vendor statement. The cloud-API page says "Usage is billed per token." and "Credits are added from your dashboard" (cloud API). The vendor's GLM-5.2 launch post (June 22, 2026) says "You also get $50 credits to test the model performance"; no account was created, so that offer was not tried
  • Key format: not recorded (apiKeyFormat is null)

Quick Start​

1. Get an API key​

  1. Visit https://subconscious.dev/platform and sign up or sign in (the quickstart describes signing up on the platform and generating an API key from the dashboard)
  2. Create an API key at https://subconscious.dev/platform/api-keys and copy it (cloud API says API keys are created and managed from the dashboard)
  3. Pricing is at https://www.subconscious.dev/pricing (monthly token plans and usage-based per-token pricing)
  4. Set SUBCONSCIOUS_API_KEY in your .env file

2. Configure​

export SUBCONSCIOUS_API_KEY=your-api-key
export SUBCONSCIOUS_MODEL=subconscious/glm-5.3-marathon   # optional — overrides the default model
export SUBCONSCIOUS_BASE_URL=https://api.subconscious.dev/v1   # optional — proxy or self-hosted gateway

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: "subconscious",
  model: "subconscious/glm-5.3-marathon",
});

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

Per-request credentials:

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

Models​

ModelVision (placeholder)$/M in · outNotes
subconscious/glm-5.3-marathonno$1.40 / $4.40 (cached $0.26)NeuroLink default. The pricing page lists it under the name "GLM-5.3 Marathon", with the description "Open-source frontier coding model".
subconscious/deepseek-v4.1-flash-marathonnonot attached (see below)NeuroLink fallback. The homepage FAQ says "On our managed inference service, we serve GLM 5.3 Marathon and DeepSeek V4.1 Flash Marathon." The pricing page lists a different id, subconscious/deepseek-v4-flash-marathon, with the description "Multimodal & high-throughput DeepSeek V4". Not probed with a key.
subconscious/tim-qwen3.6-27bnonot attachedNeuroLink fallback. The vendor's May 21, 2026 blog post "Introducing TIM-Qwen3.6-27B" says the model is "available via the OpenAI chat completions and Anthropic messages API formats". Not probed with a key.

The three ids above are on the unauthenticated GET /v1/models roster (2026-09-29). The vision value (false, shown as no in the table) of the three listed models is a schema-required placeholder the vendor does not publish per model.

Pricing comes from the pricing page (https://www.subconscious.dev/pricing, retrieved 2026-09-29), in USD per 1M tokens. It is attached to subconscious/glm-5.3-marathon, the id the page lists. The page lists subconscious/deepseek-v4-flash-marathon at $0.14 input / $0.28 output / $0.0028 cached; that id differs from the roster id subconscious/deepseek-v4.1-flash-marathon, so no pricing is attached to the roster id.

models.defaultContextWindow (200,000) and models.defaultMaxOutputTokens (4,096) are placeholders the vendor does not publish per model; the catalog schema requires both. Neither was probed.

The roster also lists test-dsv4-vision, which is not in this catalog.

Fallback order: subconscious/deepseek-v4.1-flash-marathon → subconscious/tim-qwen3.6-27b.

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 Subconscious — docs- and roster-verified, not live-verified:

ProbeResult
Rosterunauthenticated GET https://api.subconscious.dev/v1/models, HTTP 200, 4 models, 2026-09-29 — no key used
Route checkunauthenticated GET https://api.subconscious.dev/v1/chat/completions answered HTTP 405 with allow: POST, 2026-09-29; no POST was sent
Billingpublic pages read without an account: the cloud API page and the pricing page. The GLM-5.2 launch post (June 22, 2026) says "You also get $50 credits to test the model performance"; no account was created, so that offer was not tried, 2026-09-29
Error statusesthe OpenAPI spec lists 401 "Unauthorized because of a missing or invalid API key" and 429 "Rate limited due to too many requests or tokens per minute"; errorRules match on status alone and were not probed live
Tools / structured outputdeclared false; nothing was exercised live
Live capability sweepnot run. evidence.liveMatrix is null. To run it: npx tsx test/continuous-test-suite-provider-matrix.ts --provider=subconscious with a real key, then record the result.

Error statuses​

The OpenAPI spec (https://docs.subconscious.dev/api-reference/openapi.json) lists these responses for /chat/completions:

StatusVendor descriptionNeuroLink class
400Bad request due to invalid parameters or unsupported optionsnot in errorRules
401Unauthorized because of a missing or invalid API keyauthentication
429Rate limited due to too many requests or tokens per minuterate-limit
500Internal server errornot in errorRules

See also​