Midium Plus
Unlimited vaults. Lite cloud agents. One subscription.
Midium Desktop is free with one personal vault. Plus unlocks as many vaults as you need, folders linked from anywhere on your Mac, and the Lite Flex catalog in Desktop, Playground, and Agent — no token wallet, no usage limits. $18/mo.
Personal
One flat-rate subscription with unlimited vaults and unlimited cloud usage
Need more vaults? Not enough compute for self-hosting? Midium Plus includes unlimited vaults and unlimited use of any Lite Agents in the cloud catalog. Each vault can index folders, codebases, and cloud storage — link what you already have without moving it.
Sharing with a team? Get a Hub license.
Midium Plus
Every Mac you sign in to
$18 / month
First 7 days free
- Unlimited Vaults
- Linked folders
- Cloud storage copies
- Chats across Vaults
- Lite Flex catalog, no usage limits
What's included
Your personal stack, covered.
One subscription for vaults, linked folders, and Lite cloud agents. Nothing team-shared — that's Teams.
Unlimited Lite agents on Flex
Use Midium’s Lite Flex catalog in Desktop, Playground, and Agent — no token wallet, no usage limits.
As many Vaults as you want
Work, personal notes and each project in its own Vault.
Link folders where they are
Add Documents, Desktop or a repo to a Vault without moving a file.
Ask across Vaults
Ground a chat in several Vaults, or all of them, and get one answer.
Lite Flex catalog
The cloud agents included with Plus.
Live list of models on the Midium Plus Flex catalog — Lite agents, plus embeddings and System One when they're routed. Use them in Desktop, Playground, and Agent with no usage limits.
11 models
Midium Plus
Lite Agents
Flint
balanced · text-text
Flint is a new MoE coding agent by a Western lab
- Quant
- —
- Weights
- —
- Context
- —
Gemma 4 12B
balanced · text-text
Gemma 4 12B is one of the best agents for it's size. Great for chat and research
- Quant
- —
- Weights
- —
- Context
- —
Gemma 4 26B A4B
balanced · text-text
A fast and useful general purpose agent that is great for chats and research
- Quant
- —
- Weights
- —
- Context
- 256K context
Gemma 4 31B
balanced · text-text
Slower than Gemma 4 26B A4B, but has strong benchmarks and reasoning capability
- Quant
- —
- Weights
- —
- Context
- —
Laguna S 2.1
balanced · text-text
A fast and competent coding agent developed in the US by Poolside
- Quant
- —
- Weights
- —
- Context
- —
Qwen3.6 35B A3B
balanced · text-text
One of the best coding agents for it's size, as an MoE it is incredibly fast
- Quant
- —
- Weights
- —
- Context
- —
Qwen3.8 27B
balanced · text-text
The best coding agent in the Lite class of models
- Quant
- —
- Weights
- —
- Context
- 262K context
Embeddings
BGE-M3 (BAAI)
embedding · embedding
BGE M3 Embeddings
- Quant
- —
- Weights
- —
- Context
- 512 context
Qwen3 Embedding 8B
embedding · embedding
Qwen's Embeddings model
- Quant
- —
- Weights
- —
- Context
- —
System One
KnowLine 26B-A4B Gen2
system_one · system_one
A powerful and benchmark leading System One decisions model
- Quant
- —
- Weights
- —
- Context
- —
Perplexity Decider Flash v1.2
system_one · system_one
Perplexity's Flash Decider model, version 1.2
- Quant
- —
- Weights
- —
- Context
- —
