gpubox.ai

GPUBox for developers

Sovereign, OpenAI-compatible inference.
Billed in £.

Point your existing OpenAI client at our gateway and ship. UK data residency, transparent per-call pricing in pounds. Built for AI builders who want a sovereign, auditable answer — not another opaque endpoint.

Get an API keyLive API reference →
swap the base_url — keep your OpenAI code
from openai import OpenAI

client = OpenAI(
    api_key="gpb_...",
    base_url="https://api.gpubox.ai/v1",
)
# everything below is unchanged from your OpenAI code

Capabilities

OpenAI-compatible surfaces.

Read the docs →

Chat completions

live

Qwen2.5-32B-Instruct and QwQ-32B reasoning. Streaming, tool use, JSON mode. Drop-in /v1/chat/completions.

/v1/chat/completions

Vision / image input

live

Qwen2.5-VL-7B-Instruct. Send images inline as image_url (base64 data-URI or https URL) on the same /v1/chat/completions endpoint. OCR, screenshot/UI analysis, charts, visual Q&A.

/v1/chat/completions

Embeddings

live

BGE-M3 multilingual 1024-dim dense vectors, 8k context, L2-normalised. /v1/embeddings.

/v1/embeddings

Audio / speech-to-text

live

Whisper-large-v3-turbo via faster-whisper. 100+ languages, verbose JSON with segment timestamps. /v1/audio/transcriptions.

/v1/audio/transcriptions

Images

live

Text-to-image generation on UK GPUs, OpenAI-compatible request shape. Live on the API today (FLUX).

/v1/images/generations

Fine-tunes (LoRA)

coming soon

Train a LoRA adapter on your data, host it with us, call it by name — weights stay in the UK. Coming with the Factory product; not yet on the API.

/v1/training/runs

Compliance & residency

UK GDPR, stated plainly.

Operated by a UK-incorporated counterparty (Mobile Paradigm Consultancy Ltd, VAT GB397067846). UK GDPR Article 28 DPA with named-subprocessor disclosure and IDTA / SCCs where data flows outside the UK. Per-call audit log retained 30 days minimum.

Residency

UK data residency; billing and payment rails in £. Inference runs on UK-domiciled hardware — we never claim compute or data residency outside the UK.

Workload fit

Best fit for fine-tunes, batch jobs, evaluation, and development inference. We do not position GPUBox for real-time, low-latency consumer serving.

SLA

B2B SLA is best-effort and lower than our consumer tier — no uptime percentage is promised today. Dedicated-capacity SLAs are available on enterprise contracts.

Pricing — GBP

Published pay-as-you-go rates.

Chat completions (LLM)

£1.00per 1M tokens

Speech-to-text

£0.005per audio minute

Embeddings

£0.05per 1M tokens

Published pay-as-you-go rates, excluding VAT. UK customers see VAT added at checkout. See full pricing for the complete card.

See full pricing →

Talk to us about your workload

Planning a fine-tune, a batch run, or something with unusual shape? Tell us about it and we will follow up. This is a conversation, not a gate — you do not need to wait for us to start building.

Just want to start building? You don't need to ask. Sign up and we issue your API key on first sign-in — no invite, no card up front.

Sign up free →
Send workload details →

Enter your name and company to continue. No API key is issued and no account is created from this page.