Execute your workflows with Nace Small Models
One API key for Drex, which puts a probability on every call, and NDI, which reads 36 file types and cites the source of every answer.
Works with
Leading performance at a fraction of the cost.
Our most accurate setting reads documents as well as the best tool we tested, for less than half the price. Our lightest setting is the cheapest of them all and still beats all but one of the leading AI models.
Parsing leaderboard export, 15 September 2026. Cost is USD per 1,000 pages as measured. ParseBench is scored on its main metric, Content Faithfulness, only.
Builders on Drex, in their words.
V
Vishesh Baghel@VisheshBaghell · Sep 25
$499,235 of unbilled work across 1000 work orders. @NaceAI's drex found $482,260.
drex reads the note and the invoice and flags what's missing. code does the math. when drex isn't sure, a human decides.
I
Ivan Fioravanti@ivanfioravanti · Sep 25
Drex: faster, cheaper, better than jev, 32k context and with open weights coming soon? I will surely give it a try!
I’m using Drex as a fast decision layer inside my project Elyreon/NOVA mainly for routing tasks, choosing tools/models, classifying failures, retries, urgency, and handoffs.
the bot already responds sub 1s so i cant imagine how much better drex will make it
I
Israfil@Israfilv2 · Sep 30
Nace shows Drex playing a strategy game. i gave it a harder opponent: my timeline
Spoiler Guard sends each post to the new decision model and blurs the spoilers, 352 post checked in one session with sub-second calls
Using a massive LLM to decide which tool to call feels like hiring a CEO to sort your inbox.
Drex targets those decisions: routing, ranking, classification and guardrails.
Switching was easy. I pointed my existing Jev setup at Drex, kept the same SDK, changed 3 environment variables, and it worked.
I see a good fit for agent routing, tool selection, document classification, and choosing the next action in browser automation.
A
ahmadghoOct 7
i was using jev for playing vs a bot in an old game called Little fighter 2
drex 1.5 is actually very comparable with jev's behaviour in game infact might be better
also it's less latent and very fast i've dropped jev eversince and been optimizing for drex
D
Dean W. Perkins@deanwperkins · Oct 6
someone wired a small open model to a real security camera
389 ms per call
the vision model sees, drex decides
128k context means it can route with real state instead of tiny summaries.
the interesting part isn’t more generation. it’s better decisions before generation starts.
See it work from a terminal.
An agent with a Console key reads an AP packet, finds the page that matters and decides what to do with it. Every call in the window is a Console endpoint, and every value comes back with the page it was read from.
Everything your agent needs, behind one key.
Authorization: Bearer $NACE_API_KEY
One key, one credit and one error envelope for both series.
Decision
Drex puts a calibrated probability on every option, in one forward pass.
Explore DrexPOST /v1/systemoneINV-88214
state
Invoice INV-88214 bills 1,200 units at $14.10. PO-5512 authorised 1,200 at $13.90. Goods receipt confirms 1,200 delivered, 2 days late.
choicedisposition
- hold_for_buyer0.612
- approve0.301
- dispute0.087
a probability per option
noulwithin_price_tolerance
P(yes)0.148
one probability of yes
scorevariance_severity
- 0.214low
- 0.583moderate
- 0.171elevated
- 0.032high
a distribution over ordered levels
Perception
Nace Document Intelligence reads any file and points every value back to its source.
Explore NDIPOST /v1/documents/{task}invoice_2023T210.pdf
Parse
# Invoice 2023T210
Item Amount Bookkeeping 600.00 Total 600.00 Markdown and layout blocks
Split
- Invoicep. 1–3
- Amendmentp. 4–16
- W-9p. 17
Page ranges per document
Classify
- invoice0.98
- receipt
- bank_statement
- w9
A label from your taxonomy
Extract
- /invoice_number2023T210found
- /total600.00found
- /po_numbernullnot found
Fields, found or not found
Ground
Total 600.00
p. 1
The page and box it came from
36 file types: PDF, scans, Word, PowerPoint, Excel, CSV, email, HTML, images, audio, video
Read by OCR, transcription and NER models.
Works where your agent works.
Console is plain HTTPS and JSON. Any harness that can call a tool can call it; these are the ones we test with.
Three ways in.
One platform. Four boundaries.
Same models, same API, same serving stack at every tier. The only thing that changes is the boundary around your data, from our cloud to your metal.
Your data trains your models. The weights, adapters and eval sets stay yours, inside the boundary you pick.
- Where it runs
- Our cloud
- Network path
- Public TLS
- GPU tenancy
- Shared
- Updates
- Continuous
Managed Cloud
Start on our GPUs, behind our API, with a key from Console.
- Shared GPU pool with strict tenant isolation
- Public API over TLS, no training on your data
- Usage-based pricing, no infrastructure to run
For: pilots, evaluations, mid-market teams moving fast.
- Where it runs
- Our cloud
- Network path
- Private link
- GPU tenancy
- Dedicated
- Updates
- Continuous
Dedicated Cloud
Your own GPU cluster in our cloud, reached over private links instead of the public internet.
- Single-tenant compute, pinned to your region
- PrivateLink or VPC peering into your network
- Customer-managed keys, dedicated throughput
For: regulated teams that are cloud-first but done with shared infrastructure.
- Where it runs
- Your cloud account
- Network path
- Inside your VPC
- GPU tenancy
- Dedicated
- Updates
- Pull from registry
Your Cloud
The platform installs inside your AWS or Azure account. Data stays within your boundary.
- Runs in your VPC with your keys, your logs, your IAM
- Updates pull from our registry; nothing flows back out
- The deployment model our audit and banking customers run in production
For: banks, Big 4 firms, anyone whose security review starts with “show us the network diagram.”
Talk to sales
- Where it runs
- Your datacenter
- Network path
- None
- GPU tenancy
- Dedicated
- Updates
- Signed offline bundles
On-Prem & Air-Gapped
Your datacenter, your hardware, zero outbound.
- No egress, no telemetry, offline license activation
- Signed offline update bundles on your schedule
- Ships as software on your servers or as a pre-configured appliance
For: government, defense, and sovereign environments where the network cable isn’t there.
Talk to sales
Questions
First request today. Your own cloud when you need it.
Create a key in Console and send your first request in a minute. When the data cannot leave your network, we bring the platform inside it.