LLM Tech Provider Guide

Verification status: this entry is docs- and roster-verified only — not yet live-verified. It was onboarded credential-free: no LLM Tech API key was ever created or used. Every field below comes from LLM Tech's public model page (https://llmtech.eu/models/qwen3.8-27b) and an unauthenticated GET /v1/models call (no auth header, no key). No POST request was ever sent, so no generate, stream, tool-calling or structured-output capability has been exercised end to end. Treat capabilities.tools, .structuredOutput and .thinking as documentation-grade, not wire-proven, until evidence.liveMatrix is filled in by a live run.

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

Key Facts

Provider id: llmtech
Protocol: OpenAI-compatible (/v1/chat/completions). The vendor's own quickstart uses the OpenAI SDK's client.chat.completions.create(...) against this base URL.
Base URL: https://api.llmtech.eu/v1
Default model: nvidia/Qwen3.8-27B-NVFP4
Models in catalog: 1 — LLM Tech's entire hosted roster, confirmed by an unauthenticated GET /v1/models (HTTP 200, exactly 1 model) re-run on 2026-09-28.
Hosting: self-hosted on EU hardware — GPU node in Italy, edge network in Nuremberg.
Streaming: documented (SSE, OpenAI format, stream=True in the vendor's sample) — not live-probed.
Tool calling: documented ("yes, with structured outputs") and the model's own features array on the live /v1/models response lists tools — not live-probed.
Structured output: the model's features array lists json_mode and structured_outputs; the vendor's docs mention structured outputs only in the same tool-calling sentence, with no separate response_format wire example — not live-probed. Combining tools with a JSON schema in one request is not documented either way, so structuredOutputWithTools is declared false until a combined probe is run.
Embeddings: not documented anywhere on the vendor's site.
Vision: real image support, not a text-only model despite some third-party listings — the vendor's page states the vision tower is kept in BF16 and works; image tokens count toward prompt_tokens and bill at the input rate (e.g. a 64×64 PNG costs 64 image tokens), and images must go in a user message, not a system message. The live roster's features array also lists vision. Not live-probed.
Reasoning / thinking: controllable via enable_thinking and reasoning_effort request parameters; the live roster's features array lists reasoning.
Billing: pay-per-token USD, no free tier documented — "no subscription, no minimums." Prompt caching is automatic and billed at the cache rate.
Signup: no public signup or self-service key page. Self-service is described as "planned." Keys are issued manually by emailing [email protected]; the account dashboard at https://llmtech.eu/cabinet/ confirms this model — "your key is your login," "we hold no passwords and no accounts" — it is a billing/usage view for an existing key, not a signup form.
Key format: none declared.

Quick Start

1. Get an API key

LLM Tech does not publish a public signup or self-service key-issuance flow:

Email [email protected] to request a key (self-service is planned but not live as of 2026-09-28).
Once issued, paste the key into https://llmtech.eu/cabinet/ to see current-month billing, a daily usage chart, and invoice line items — there is no separate account or login.
Confirm current pricing before heavy use: input $0.25/M, output $2.09/M, cached input $0.04/M tokens, per https://llmtech.eu/models/qwen3.8-27b.
Set LLMTECH_API_KEY in your .env file.

2. Configure

export LLMTECH_API_KEY=your-api-key
export LLMTECH_MODEL=nvidia/Qwen3.8-27B-NVFP4      # optional — overrides the default model
export LLMTECH_BASE_URL=https://api.llmtech.eu/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 NVFP4 quantization in one paragraph." },
  provider: "llmtech",
  model: "nvidia/Qwen3.8-27B-NVFP4",
});

console.log(result.content);

# CLI
npx @juspay/neurolink generate "Hello" --provider llmtech

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

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

Models

Model	Context	Vision	$/M in · out	Notes
`nvidia/Qwen3.8-27B-NVFP4` ⭐	262,144	yes	$0.25 / $2.09 (cached $0.04)	LLM Tech's only model. Reasoning-capable via `enable_thinking`/`reasoning_effort`. Max output 32,768 tokens.

Context window and max output shown are the catalog defaults (262,144 / 32,768 tokens), taken directly from the model's own entry in the live, unauthenticated GET /v1/models response (context_length, max_output / max_output_tokens) on 2026-09-28. Pricing is that same response's pricing_per_m fields, which match the vendor's model page exactly.

Fallback order: none — nvidia/Qwen3.8-27B-NVFP4 is the only model LLM Tech serves, so it is both the default and its own (schema-exempt) single-entry fallback list.

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 records for LLM Tech — and, just as importantly, what it does not yet record:

Probe	Result
Roster	unauthenticated GET `/v1/models`, HTTP 200, 1 model, 2026-09-28. No API key was used or required for this call.
Auth rejection	Not probed. Would require a `POST` with an invalid key, which this credential-free onboarding pass does not send.
Live capability sweep	Not run. `evidence.liveMatrix` is `null`. `capabilities.tools`, `.structuredOutput`, `.thinking` and vision are set from the vendor's model page plus the served model's own `features` array on the live roster — not from an executed chat/stream/tool-call/schema request. `structuredOutputWithTools` and `toolsWithStreaming` are `false` because no combined request was — or could be — attempted without a key.

Until a live run fills in evidence.liveMatrix, treat this provider as documentation-grade: the wire shape (base URL, OpenAI-compatible request/ response field names) is well-documented, but no NeuroLink call has actually exercised it.

Troubleshooting

Symptom	Cause	Fix
`Invalid LLM Tech API key`	`LLMTECH_API_KEY` unset or wrong	LLM Tech's docs do not publish an example auth-error body or code string; re-check the key with your issuer
Model not found	The roster changed since 2026-09-28	Re-check with an unauthenticated `GET https://api.llmtech.eu/v1/models`
No signup page found	Expected — LLM Tech has no public signup page yet	Email `[email protected]` for a key; self-service is planned but not live
Unexpectedly high image cost	Image tokens bill at the input rate	Budget accordingly — a 64×64 PNG costs 64 image tokens per the vendor's own example

Key Facts​

Quick Start​

1. Get an API key​

2. Configure​

3. Use it​

Models​

Verification status​

Troubleshooting​

See also​