Munito AI
Munito AI provides open-model inference for Europe through fast, OpenAI- and Anthropic-compatible APIs. It aims to let companies use independent open models with EU processing and storage, privacy by default, and no need to manage GPUs.
At a Glance
- Enterprise AI software companies
- Developers and startups building agents and products on open models
- Finance, healthcare and legal industries
- European organizations with GDPR, data-residency or sovereignty requirements
- +1 more
AI Tools by Munito AI
(1)
Munito AI
Open Model Inference API
Discussions
No discussions yet
Be the first to start a discussion about Munito AI
Latest News
Decision model pplx-decider-v1.2 expanded to a 262,144-token state and decision models became generally available rather than preview.
LightOnOCR-3-4B replaced LightOnOCR-2-1B in the EU and US OCR service.
Munito added DeepSeek Harness and Prime Agent integration guides and expanded OCR table extraction.
OpenCode began listing Munito AI and Munito AI (EU) providers; Munito also published provider packages/guides for pi, OpenClaw and Hermes Agent.
Products & Services
Open-model inference over OpenAI-compatible Chat Completions, Responses and speech-to-text endpoints plus Anthropic-compatible Messages. Supports global, EU and US hosts, model choice, tool calling, structured output, reasoning and streaming.
Converts PDFs, scans and handwritten images to Markdown; supports table extraction and invoice-to-structured-data workflows, with EU or US processing and per-page billing.
Preview API for isolated Linux microVMs with their own kernel, disk and network; agents and code can execute commands and read/write files in a sandbox.
Contracted open-model inference with volume rates, monthly invoicing and enterprise terms.
Market Position
Munito positions itself as an independent, Europe-focused open-model inference provider rather than a cloud VM vendor or a model lab serving only its own models. Its differentiation is EU-first processing and storage, privacy by default, compatibility with major OpenAI and Anthropic SDKs/agents, model choice across multiple open-model families, and a unified API spanning text, speech, documents and sandbox compute. The relevant competitive set includes European/regional inference and AI-endpoint providers such as OVHcloud AI Endpoints, Scaleway, Hugging Face Inference Providers, Together AI, Fireworks AI and Groq.
Leadership
Founders
Alexey Eryshev
Founder and CEO; previously led inference at Mistral AI, where his team built the API and inference products and he built Le Chat's Deep Research. Before Mistral, he spent five years at Airbnb building Hosting Services from scratch into a global product; he has also worked at Meta.
Executive Team
Alexey Eryshev
Founder and CEO; President of Munito AI France
Former inference lead at Mistral AI; previously spent five years at Airbnb building Hosting Services, and has worked at Meta.
Founding Story
Alexey Eryshev founded Munito after seeing at Mistral AI that Europe lacked well-established independent providers able to run open-model inference at scale. His initial vision was to offer the best open models, run in Europe, behind an API developers already know, giving companies privacy, regional control and sovereignty without operating their own GPUs.
Business Model
Revenue Model
Usage-based inference and OCR. Serverless customers use prepaid credits and pay posted prices per million tokens or per 1,000 OCR pages; higher-volume customers can use committed-spend contracts with monthly invoicing and volume rates, while dedicated endpoints are contracted reserved GPU capacity.
Pricing Tiers
New accounts with a verified phone number can use the free tier subject to capacity; an organization returns to the free tier when its paid balance reaches zero.
Standard service tier. For example, Qwen3.8-27B is $0.20 per million input tokens and $2.70 per million output tokens in the EU and US.
Higher-cost service tier intended for priority service; prices vary by model and region.
Lower-cost service tier; prices vary by model and region.
OCR is billed per page read, with price determined by model, region and service tier.
Target Markets
- Enterprise AI software companies
- Developers and startups building agents and products on open models
- Finance, healthcare and legal industries
- European organizations with GDPR, data-residency or sovereignty requirements
- Professional and business users needing hosted inference, OCR or dedicated capacity
- AI products and agents that need EU inference and storage
- Sensitive enterprise workloads in finance, healthcare and legal services
- Developers migrating existing OpenAI or Anthropic SDK applications to open models
- Coding agents and autonomous agents that need tool calling or an isolated computer
- PDF, scan, handwriting, table and invoice extraction
- Companies that need regional control and privacy without managing GPUs