Skip to main content

Bytez Provider Guide

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

Verification status: this entry is docs-verified only, not yet live-verified. Bytez's GET /models needs a key, so the model ids come from its public docs and the roster has not been checked. The fields below come from Bytez's own public documentation and unauthenticated GET requests — no account was created and no API key was used to build it. evidence.liveMatrix is null until someone runs the live capability matrix with a real key (see Verification status below).

Key Facts​

  • Provider id: bytez
  • Protocol: OpenAI-compatible (/chat/completions) — the docs' OpenAI-client examples create a client with this base URL and call chat.completions.create (Chat Completions page)
  • Base URL: https://api.bytez.com/models/v2/openai/v1
  • Default model: Qwen/Qwen3-4B (NeuroLink default)
  • Models in catalog: 2 models named in the Bytez docs
  • Streaming: supported — the "Streaming (Open Source)" example sets stream: true and reads event.choices[0].delta.content (Chat Completions page)
  • Request fields listed in the API reference: model, messages, max_completion_tokens (default 256), temperature (default 0.7), stream, top_p, presence_penalty, frequency_penalty, logprobs and top_logprobs (API reference). The examples on the Chat Completions page send max_tokens
  • Tool calling: false — declared in this entry and not exercised
  • Tools while streaming: false
  • Structured output: false — declared in this entry and not exercised
  • Structured output + tools together: false — not exercised
  • Embeddings: not declared on this catalog entry (false)
  • Thinking: not declared (false)
  • Billing: free-tier — the Billing & Credits page lists a Free plan, "$0 / month - Get $1 in free credits" (Billing & Credits)

Quick Start​

1. Get an API key​

  1. Open the API Dashboard at https://bytez.com/api; the Get started page (https://docs.bytez.com/model-api/docs/get-started.md) says "Copy your key from the API Dashboard"
  2. Free plan (https://docs.bytez.com/model-api/docs/billing.md): "$0 / month - Get $1 in free credits", with "Run open models up to 7B parameters" and "Credits refresh every 4 weeks"; the Billing Cycle section lists Billing: None for the Free Plan
  3. The same page lists a Pay-as-you-go plan with "Run open models up to 120B parameters" and "Add credits anytime"
  4. Set BYTEZ_API_KEY in your .env file

2. Configure​

export BYTEZ_API_KEY=your-api-key
export BYTEZ_MODEL=Qwen/Qwen3-4B   # optional — overrides the default model
export BYTEZ_BASE_URL=https://api.bytez.com/models/v2/openai/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: "bytez",
  model: "Qwen/Qwen3-4B",
});

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

Per-request credentials:

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

Models​

ModelContextVisionPricingNotes
Qwen/Qwen3-4B ⭐not setnoper second of inferenceNeuroLink default. The model in the "Basic usage (Open Source)" and "Streaming (Open Source)" examples of the Chat Completions page and in the Basic usage example of the chat Get started page.
Qwen/Qwen3-1.7Bnot setnonot setNeuroLink fallback. Named in the API reference's model field description: "The ID of the model to run (e.g., Qwen/Qwen3-1.7B, openai/gpt-4)".

Model ids are copied from these pages, retrieved 2026-09-29:

Open models are billed per second of inference (the "How Credits Work" table on the Billing & Credits page). That page's "Pricing by Model Size" table lists 7B at $0.000072 per second, 15B at $0.000108, 35B at $0.000144, 70B at $0.000216 and 120B at $0.00036. The catalog stores no per-token price, because pricingPerMTok is a per-token figure.

models.defaultContextWindow (8,192) and models.defaultMaxOutputTokens (1,024) are placeholders the vendor does not publish. The per-model context and output figures are left unset.

Vision: the catalog sets vision: false for both models. The Chat Completions API reference types message content as string.

Fallback: fallbacks lists Qwen/Qwen3-1.7B, which is the runtime fallback model name the loader derives.

Not in the catalog: closed-source provider models (for example openai/gpt-4) are addressed with a provider prefix, and their curl examples on the Chat Completions page add a provider-key header. The catalog schema has no field for an extra request header, so none are listed.

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 Bytez — docs-verified only, not live-verified:

ProbeResult
Rosterunauthenticated GET https://api.bytez.com/models/v2/openai/v1/models answered HTTP 401 with the body {"error":"Unauthorized"}, 2026-09-29 — no key used. The model ids come from the docs pages listed under Models; the roster has not been checked
Auth-failure shapeunauthenticated GET of both .../v1/models and .../v1/chat/completions answered 401 with {"error":"Unauthorized"}, 2026-09-29. The Models API reference describes 401 as "Auth error - check your api key and how you're sending it." (page). errorRules carry 401, 402 and 429
Auth headerthe curl examples on the Chat Completions page send Authorization: BYTEZ_KEY, and its OpenAI-client examples pass the key as apiKey. The Get started page has the code comment # add an Authorization header, with value "Key {BYTEZ_KEY}" above a curl example that sends -H "Authorization: BYTEZ_KEY". NeuroLink's OpenAI-compatible base sends Authorization: Bearer <key> by default. That request was not exercised — it needs a key
BillingBilling & Credits lists a Free plan, "$0 / month - Get $1 in free credits", and its Billing Cycle section lists Billing: None for the Free Plan, 2026-09-29
Tools / structured outputdeclared false (see Key Facts for the request fields the API reference lists). Neither was exercised, and no combined tools+schema request was sent — structuredOutputWithTools stays false
Live capability sweepnot run. evidence.liveMatrix is null. The command is npx tsx test/continuous-test-suite-provider-matrix.ts --provider=bytez, and it needs a real key.

This entry is not live-verified until evidence.liveMatrix is filled in.

Documented error responses​

StatusWhat the Bytez docs say
401The Models API reference describes 401 as "Auth error - check your api key and how you're sending it." (page)
402The Billing & Credits page shows the response for an account whose credits ran out with auto-reload disabled, with the message "Insufficient credits. Please add credits to continue." (page)
429The Models API reference documents 429 as "Too Many Requests – You have hit the rate limit."; the Billing & Credits page says requests over the open model rate limits "are rejected with a rate-limit error"

See also​