Skip to main content

Parasail Provider Guide

Parasail is a Tier-2 catalog provider: its integration is one JSON file (src/lib/providers/catalog/parasail.json) rather than hand-written code. That file is the source of truth for this page. The Authentication page says "The API is fully OpenAI-compatible."

Verification status: this entry is docs-verified only, not yet live-verified. Parasail's GET https://api.parasail.io/v1/models needs a key (an unauthenticated GET on 2026-09-29 answered HTTP 401), so the model ids come from Parasail's public docs (Models page, retrieved 2026-09-29) and the roster has not been checked. No account was created and no API key was used to build this entry. evidence.liveMatrix is null until someone runs the live capability matrix with a real key (see Verification status below).

Key Facts​

  • Provider id: parasail
  • Protocol: OpenAI-compatible (/chat/completions) — the Authentication page says "All Parasail API requests use bearer token authentication. The API is fully OpenAI-compatible."
  • Base URL: https://api.parasail.io/v1
  • Default model: parasail-llama-33-70b-fp8 (NeuroLink default)
  • Models in catalog: 12 (curated from the 33 rows of the Models page)
  • Streaming: supported — the Chat Completions reference has a Streaming section describing stream: true with server-sent events
  • Tool calling: model-dependent — the Tool/Function Calling guide lists the serverless models that return structured tool calls in a table and names other models that "don't support the tools parameter today"
  • Tools while streaming: not declared (false) — no combined request was sent
  • Structured output: supported — the Structured Output guide documents response_format with type json_schema and guided_json, and lists models in a table with a "Guided JSON / JSON Schema" column
  • Structured output + tools together: not declared (false) — no combined probe was possible without credentials, so NeuroLink omits response_format automatically whenever tools are attached
  • Embeddings: not declared on this catalog entry. The Chat Completions reference's endpoint table lists POST /v1/embeddings, but NeuroLink's generic ConfiguredOpenAICompatProvider (the class Tier-2 catalog entries use) does not implement embed()/embedMany(), so this flag tracks that, not the vendor's own API surface
  • Thinking: not declared — the entry sets no reasoning request parameter, although the Model-specific Notes page documents model-specific thinking and reasoning controls
  • Billing: recorded as no-free-tier, which is the schema placeholder (the catalog schema has no unknown value), not a vendor statement of signup terms — see Billing below
  • Key format: none declared

Quick Start​

1. Get an API key​

  1. Visit: https://www.saas.parasail.io/keys — the link the Authentication page gives for getting an API key; sign-in was not attempted while building this entry
  2. Create an API key and copy it when it is shown — the Authentication page steps read "Click Create API Key." and "Copy the key immediately—it's displayed only once."
  3. Billing as the vendor states it: the Pricing page says "To use models, create API keys, and access paid platform services, add a credit card. Without a card, you can still view the platform but can't use models or API keys."
  4. Set PARASAIL_API_KEY in your .env file (the docs' examples use the environment variable PARASAIL_API_KEY)

2. Configure​

export PARASAIL_API_KEY=your-api-key
export PARASAIL_MODEL=parasail-llama-33-70b-fp8   # optional — overrides the default model
export PARASAIL_BASE_URL=https://api.parasail.io/v1   # optional

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: "parasail",
  model: "parasail-llama-33-70b-fp8",
});

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

Per-request credentials work as they do for other NeuroLink providers:

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

Billing​

The Pricing page says "To use models, create API keys, and access paid platform services, add a credit card. Without a card, you can still view the platform but can't use models or API keys." The Limits and Quotas page has a rate-limit table with a product class "Serverless - Free" (RPM column "5"). The catalog's billingPolicy field takes one of three values and holds no-free-tier here as a placeholder; it is not a vendor statement of signup terms.

Models​

ModelContextMax outputVision$/M in · out (cached)Notes
parasail-llama-33-70b-fp8 ⭐131,07216,384no$0.22 / $0.50 ($0.11)NeuroLink default. Chat Completions guide example; Tools ✅ and Tool Choice ✅; Y under Guided JSON / JSON Schema, Regex and Choice.
parasail-llama-4-maverick-instruct-fp8524,28832,768no$0.35 / $1.00 ($0.17)Tools ✅ and Tool Choice ✅; Y under Guided JSON / JSON Schema, Regex and Choice. Fallback.
parasail-gpt-oss-120b131,072131,072no$0.10 / $0.75 ($0.055)Tools ✅ and Tool Choice ✅. Model-specific Notes page has "GPT-OSS reasoning control" and "Structured outputs with GPT-OSS" sections. Fallback.
parasail-gpt-oss-20b131,072131,072no$0.03 / $0.15 ($0.02)Tools ✅ and Tool Choice ✅. Same Model-specific Notes sections as parasail-gpt-oss-120b. Fallback.
parasail-deepseek-v4-flash1,048,5761,048,576no$0.14 / $0.28 ($0.07)Tools ✅ and Tool Choice ✅. Fallback.
parasail-deepseek-v4-pro1,048,5761,048,576no$1.74 / $3.48 ($0.10)Tools ✅ and Tool Choice ✅. Fallback.
parasail-kimi-k31,048,5761,048,576no$3.00 / $15.00 ($0.30)Tools ✅ and Tool Choice ✅. Fallback.
parasail-glm-52262,144262,144no$1.40 / $4.40 ($0.26)Tools ✅ and Tool Choice ✅. Fallback.
parasail-qwen3p5-35b-a3b262,144262,144no$0.15 / $1.00 ($0.05)Tools ✅ and Tool Choice ✅; the Tool/Function Calling guide's Chat Completions example uses it. Fallback.
parasail-qwen25-vl-72b-instruct128,000128,000yes$0.80 / $1.00 ($0.40)The Multi-Modal guide's image example uses it, hence vision: true and models.visionModel. Tool/Function Calling guide: "also emits tool calls, but only reliably with" tool_choice="required". Fallback.
parasail-gemma3-27b-it131,072131,072no$0.08 / $0.45 ($0.04)Y under Guided JSON / JSON Schema, Regex and Choice. Tool/Function Calling guide names it among models that "don't support the tools parameter today". Fallback.
parasail-minimax-m31,048,576524,288no$0.30 / $1.20 ($0.06)Tools ✅ and Tool Choice ✅. Fallback.

Context window, max output and prices are read from the Models page (retrieved 2026-09-29), which says "This catalog is generated daily from the live models endpoint." The Pricing page shows the same input, output and cached-input figures for each of the 33 rows. The tool-calling markers come from the Tool/Function Calling guide and the Y markers from the Structured Output guide. The Multi-Modal guide is the source for the one vision: true.

models.defaultContextWindow (128,000) and models.defaultMaxOutputTokens (16,384) are placeholders the vendor does not publish as general values. They apply to model ids outside the catalog and equal the Context window cell of the Qwen2.5-VL 72B row and the Max output cell of the Llama 3.3 70B (FP8) row on the Models page. Each model's status is production, a value the catalog schema requires; it is not a vendor statement. Each vision: false is a placeholder for a boolean the schema requires, not a vendor statement.

Fallback order when the default is unavailable: parasail-llama-4-maverick-instruct-fp8 → parasail-gpt-oss-120b → parasail-gpt-oss-20b → parasail-deepseek-v4-flash → parasail-deepseek-v4-pro → parasail-kimi-k3 → parasail-glm-52 → parasail-qwen3p5-35b-a3b → parasail-qwen25-vl-72b-instruct → parasail-gemma3-27b-it → parasail-minimax-m3. The runtime fallback model name the loader derives (fallbacks[1]) is parasail-gpt-oss-120b.

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 Parasail — docs-verified only, not yet live-verified:

ProbeResult
Rosterunauthenticated GET https://api.parasail.io/v1/models answered HTTP 401 on 2026-09-29 — no key used, so the model ids are taken from the Models page (retrieved 2026-09-29) and the roster is not verified
Billingthe Pricing page says "To use models, create API keys, and access paid platform services, add a credit card. Without a card, you can still view the platform but can't use models or API keys."; the Limits and Quotas page has a "Serverless - Free" rate-limit row; no signup was performed, 2026-09-29
Auth-failure shapeunauthenticated GETs of /v1/models, /v1/chat/completions and /v1 each answered HTTP 401 with content-type application/json;charset=ISO-8859-1 and the body text "Unauthorized. No bearer token provided." (plain text, no error code field), 2026-09-29; no POST was sent. errorRules match the statuses 401, 403, 404 and 429 from the Error responses table of the Chat Completions reference
Tools / structured outputdocumented on the Tool/Function Calling and Structured Output guides; neither was exercised live, and no combined tools+schema request was sent — structuredOutputWithTools stays false
Live capability sweepnot run. evidence.liveMatrix is null. Before treating this provider as production-ready, run npx tsx test/continuous-test-suite-provider-matrix.ts --provider=parasail with a real key and record the result.

Do not treat this entry as equivalent to a live-verified Tier-2 provider (e.g. FriendliAI, Novita AI) until that live matrix has been run and evidence.liveMatrix is filled in.

Documented error statuses​

The Chat Completions reference says "Errors use OpenAI-compatible status codes and a JSON body with an error object (message, type, code)." and lists these statuses:

StatusParasail's description
400Malformed request—invalid JSON, missing required field, or unsupported parameter value
401Missing or invalid API key
403API key lacks access to the requested model or resource
404Model or endpoint not found
429Rate limit or quota exceeded—back off and retry
500Internal server error—retry with exponential backoff

The Tool/Function Calling guide says of models outside its table: "Depending on the model, a request with tools returns an error or plain text with no tool_calls."

See also​