AI software testing platform whose Aximo agent runs web and mobile tests from plain-English steps, with a free tier and credit-based plans.
Freemium
Lumai
Lumai builds Iris, optical AI inference servers that run large language models in data centers.
What is Lumai?
Lumai is an Oxford-based company building optical computing hardware for AI inference. Its Iris server family uses an optical matrix multiplier, which does the core calculations with beams of light in three-dimensional space, alongside a digital processor that handles the rest of the workload. The first server, Iris Nova, runs Llama 8B and 70B models, and research institutions, enterprises, neo-clouds and hyperscalers can evaluate it in data center environments. Iris Aura and Iris Tetra are the next servers in the family. Lumai says the design targets compute-bound steps such as prefill, can work alongside decode-focused accelerators, fits standard racks with air cooling, and aims to cut power use and total cost against silicon-only accelerators. Lumai does not publish prices; the hardware is offered through direct contact, so approach the company to arrange an evaluation.
Use cases
- Evaluate optical hardware for LLM inference
- Run Llama models in a data center
- Speed up prefill in inference pipelines
- Reduce power use for AI workloads
- Test alternatives to silicon accelerators
- Plan power-limited AI clusters
Who it's for
Why use Lumai
- Aims to lower power use for compute-bound inference compared with digital-only chips.
- Fits conventional racks without liquid cooling, according to the vendor.
- Hybrid design keeps existing software and digital workflows in use.
- Evaluation is open to enterprises and research groups as well as large cloud operators.
Lumai features
- Optical matrix multiplication up to 2048x2048
- INT4/INT8-equivalent precision
- Runs Llama 8B and 70B models on Iris Nova
- Hybrid digital and optical architecture
- Three-tier memory design for KV cache
- Standard rack, air-cooled deployment
- Designed to work with decode-optimized accelerators
How to get started with Lumai
- Contact Lumai to arrange an evaluation of Iris Nova.
- Install the server in a standard rack in your data center.
- Load a supported model such as Llama 8B or 70B.
- Run inference workloads and measure throughput and power.
- Use it alone or as a prefill processor alongside other accelerators.
Lumai limitations
- Iris Nova is offered for evaluation in data center environments
- Pricing is not published
- Iris Aura and Iris Tetra are described as next-generation servers
Lumai pricing
Lumai is sold on custom quotes, so its prices aren't published.
Prices can change and may vary by country or exclude tax. Check the official pricing page.
Lumai alternatives
The main Lumai alternatives listed on Debuly are Groq, Cerebras, Anyscale, NVIDIA Run:ai, Union.ai and Databricks. The table compares pricing and platforms; order is A–Z.
| Tool | Pricing | Starting price | Platforms |
|---|---|---|---|
| Lumai | Contact sales | — | — |
| Anyscale | Paid | $0.01, usage-based | Web, Self-hosted |
| Cerebras | Freemium | $10, usage-based | API, Web |
| Databricks | Free trial | — | Web |
| Groq | Freemium | — | Web, API |
| NVIDIA Run:ai | Contact sales | — | Web, Self-hosted, API |
| Union.ai | Free trial | $950/mo | Web, CLI, Self-hosted |
Anyscale
Managed platform for running Ray workloads, with pay-as-you-go compute and options to deploy in your own cloud.
Cerebras
Inference cloud and API running open models on wafer-scale chips, with a free tier, pay-per-token Developer plan and coding plans.
FreemiumVisit Cerebras (official site)Databricks
Data and AI platform for data engineering, warehousing, governance, dashboards and AI agents, billed by usage on AWS, Azure and GCP.
Free trialVisit Databricks (official site)
Groq
GroqCloud inference API that serves open models such as GPT-OSS and Whisper on Groq's LPU hardware, billed per token.
FreemiumVisit Groq (official site)NVIDIA Run:ai
GPU orchestration platform that schedules, shares and monitors AI workloads across on-premises and cloud clusters.
Contact salesVisit NVIDIA Run:ai (official site)
Union.ai
Managed control plane for running AI, ML and agent workloads in your own cloud, built on the open-source Flyte runtime.
Free trialVisit Union.ai (official site)
Lumai FAQ
How much does Lumai Iris cost?
Lumai does not publish prices. Iris Nova is available for evaluation, so contact the company.
What models does Iris Nova run?
It supports billion-parameter LLM workloads, with Llama 8B and 70B named.
Who is Lumai for?
Hyperscalers, neo-clouds, enterprises and research institutions evaluating inference hardware.
Is Lumai open source?
No. It is proprietary hardware, and open-source status is not described.
What are alternatives to Lumai?
Alternatives for inference hardware include NVIDIA GPUs, Groq, Cerebras, Etched and Lightmatter.
Sources & verification
Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.
Image credits
- Logo: from cdn.prod.website-files.com · fetched · Vendor trademark, used only to identify the product; removed on request
- Cover: none collected yet, so a generated tile is shown.
Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.
More in Developer platforms 4
All developer platforms toolsMicrosoft's MIT-licensed Python and .NET framework for multi-agent AI apps, now archived; new projects are pointed to Microsoft Agent Framework.
Open sourceOpen-source inference server, formerly Triton Inference Server, for serving models from TensorRT, PyTorch, ONNX and more in production.
Open sourceIntegration layer for AI agents and apps, with managed auth for 3,000+ APIs, prebuilt tools, triggers and MCP servers via Pipedream Connect.
Freemium
