Vora Cloud · private beta

Open models, one API.

Change your base URL and keep your code. Open-weight models from a European provider.

Join the waitlist Read the docs
Our server nodes behind one OpenAI-compatible API, called from your code.01 // OUR SERVERS02 // ONE API03 // YOUR CODE
// MODELS

Choose a model for the job.

Draft and summarise text, build an assistant, interpret an image, or generate one. These are starting points for choosing a model; test it with examples from your own work.

In production Free pilot

Qwen3.5 9B

qwen3.5-9b

Everyday text and longer conversations

Draft emails, rewrite and translate text, summarise supplied documents, and ask questions about a longer conversation.

Context
64K tokens
Maximum output
8K tokens
Weight format
Q4_K_M
License
Apache-2.0
Thinking modes
Off · Low · Medium · High
  • Chat
  • Streaming
  • Tool calling
  • JSON responses
  • Thinking

Keep in mind. Text only in Vora. A larger context lets more material fit; it does not guarantee every detail will be recalled correctly.

In production Free pilot

Nemotron 3.5 Bolt

nemotron-3.5-bolt

Assistants and tool workflows

Build support conversations, extract fields into JSON, and let an assistant choose actions that your application can carry out.

Context
32K tokens
Maximum output
8K tokens
Weight format
Q4_K_M
License
OpenMDW-1.1
Thinking modes
Off · Low · Medium · High
  • Chat
  • Streaming
  • Tool calling
  • JSON responses
  • Thinking

Keep in mind. Text only. Tool calling proposes an action; your application must validate and execute it. Long-context reliability is still being tested.

In production Free pilot

Gemma 4 26B-A4B

gemma-4-26b-a4b

Fast answers and image understanding

Ask about screenshots, charts and photos, explain code, and get quick replies in English, Spanish and other languages.

Context
128K tokens
Maximum output
8K tokens
Weight format
Q4_0 (QAT)
License
Apache-2.0
Thinking modes
Off · Low · Medium · High
  • Chat
  • Streaming
  • Tool calling
  • JSON responses
  • Thinking
  • Image understanding

Keep in mind. One image per request. Check extracted text and numbers; image understanding can miss details. Review answers to hard multi-step problems.

In production Free pilot

FLUX 3 Image Klein

flux-3-image-klein

Images from a prompt

Explore visual concepts, illustrations and early design directions by describing the image you want to create.

Image output
1024 × 1024
Per request
1 image
Weight format
INT8
License
FLUX Kommunity License
  • Image generation

Keep in mind. Text-to-image only in this pilot; editing and reference images are not exposed. Check lettering and fine details, and allow time for generation.

Production models are available by invitation during the beta. Context and output limits come from the Vora API catalogue; they are not the original model's theoretical maximum.

Context is the space for a whole request

Instructions, conversation history, documents, tool results and the generated output all use the same token window. Image inputs use space too, when supported. Reserve room for the answer: the context number is not an extra allowance for output, or permanent memory between requests.

Choose how much thinking a task needs

Use Off for a direct response, or select a supported thinking level for a task that needs more reasoning. More thinking can increase the wait and uses the output allowance; it does not guarantee a better answer. Your application chooses the level for each API request.

Thinking controls and API examples →
// WHY VORA CLOUD

OpenAI-compatible

Works with the SDKs and tools you already use.

A European provider

A European company building access to open models. Check each model's availability before integrating it.

Invite only

Free during the beta. Join the waitlist and we reply within a few days.

// TRY IT See models → 
curl https://api.voracompute.com/v1/chat/completions \
  -H "Authorization: Bearer $VORA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "qwen3.5-9b", "stream": true, "messages": [{"role": "user", "content": "Hello"}]}'
▸ streaming from qwen3.5-9b
Hi! How can I help your team today?
✓ done
// PARTNERS

In good company.

Turns server heat into heating for homes and businesses.

hotta.es ↗

Get early access.

Join the waitlist. We open the beta a few teams at a time.

We use these details only to reply about the beta. Privacy policy