nace-ai/

drex-max

$0.12 / 1M tokens

50% off

Drex Max is the flagship 27B multimodal decision model from Nace.AI. Given a state (text, JSON or images) and typed questions (yes/no, multiple choice, or ordinal score), it returns a calibrated probability for every option in a single forward pass, with no text generation. It is TypeSafe System One compatible, handles contexts up to 262K tokens, and ranks #1 on Decision Index 0.3.2 with 67.59, ahead of Drex v2.0 and Jev 1.13.0.

Public

Zero retention

262,144

Multimodal

Project License

api

Log in to run decisions.

Model Information

Drex Max

Drex Max is the flagship decision model from Nace.AI. You give it a state (text or JSON) and typed questions, and it returns a calibrated probability for every option in a single forward pass. It generates no text, and it never answers with an option you did not supply.

#1 on Decision Index 0.3.2 (67.59 public index), 6.3 points ahead of Drex v2.0 and 9.6 ahead of Jev 1.13.0.
27B dense, built on Qwen3.8-27B with non-causal full attention and a 255-way readout head.
Context up to 262,144 tokens. States longer than 8,192 tokens are cut down to the passages most relevant to your questions before the model reads them, and you are billed only for the tokens it reads.
TypeSafe System One compatible: the same request and response as Jev, so the TypeSafe SDK works by changing its base URL.
Multimodal: text, JSON and images. Every question on a request sees all of its images.

Question types

Type	You send	You get back
`noul`	a yes/no question, with optional `criteria.true` / `criteria.false`	`noul`, the probability of "yes"
`choice`	options as `criteria` (name → description or `null`), 1-255 options	`choice` (the most likely option), `probabilities` per option, and `confidence`
`score`	ordered levels as a `criteria` list, 1-255 levels	`score` (the expected level), `probabilities` per level, `legend`, and `confidence`

One request can ask up to 16 questions about the same state. The state is read once and billed once per request.

Example

curl https://api.deepinfra.com/v1/decisions \
  -H "Authorization: Bearer $DEEPINFRA_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nace-ai/drex-max",
    "state": {"ticket": "I was charged twice for the same order. Please refund the extra payment."},
    "questions": {
      "team": {"type": "choice", "instructions": "Which team should handle this ticket?",
               "criteria": {"billing": "Payments, charges, and refunds", "technical": "Bugs and outages", "other": "Anything else"}},
      "refund": {"type": "noul", "instructions": "Does the customer ask for a refund?"},
      "urgency": {"type": "score", "instructions": "How urgent is this ticket?",
                  "criteria": ["Routine", "Soon", "Urgent"]}
    }
  }'
copy

Images

Add images, a list of up to 4 images, each either a base64 data URL ("data:image/jpeg;base64,...") or an object {"content_type": "image/jpeg", "base64": "..."}. PNG, JPEG, WebP, GIF, BMP, TIFF, AVIF and HEIC are accepted, in a request body of up to 13 MiB; remote image URLs are not fetched.

Each image counts as 64 to about 4,000 input tokens depending on its size (larger images are scaled down to about 4 MP), is billed once per request whatever the number of questions, and is reported in usage.image_tokens. A repeated image is encoded once. Requests with images read the state whole, without the long-state selection; each question's row (images, state and question) can use the full 262,144-token context.

IMG=$(base64 -w0 receipt.jpg)
curl https://api.deepinfra.com/v1/decisions \
  -H "Authorization: Bearer $DEEPINFRA_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nace-ai/drex-max",
    "state": {"task": "expense review"},
    "images": ["data:image/jpeg;base64,'"$IMG"'"],
    "questions": {
      "total": {"type": "choice", "instructions": "What is the receipt total?",
                "criteria": {"a": "$7.70", "b": "$42.10", "c": "$3.20"}},
      "card": {"type": "noul", "instructions": "Was it paid by card?"}
    }
  }'
copy

The same API is served at https://api.deepinfra.com/v1/systemone. To use the TypeSafe Python SDK, set TYPESAFE_BASE_URL=https://api.deepinfra.com/typesafe and model="nace-ai/drex-max".

Performance

Figures reported by Nace.AI on the model card.

Decision Index 0.3.2 (public only)

Model	Public index	Knowledge & Reasoning	Language	Retrieval	Tools	Arts
Drex Max (27B)	67.59	57.2	72.2	68.8	82.3	53.3
Drex v2.0 (35B MoE)	61.25	47.9	65.3	63.0	76.5	53.9
Jev 1.13.0	57.96	53.9	59.2	55.4	75.1	39.1

The 37 public benchmarks in five areas, over 140,178 scored requests. Area columns are chance-corrected skill × 100. Unanswered requests count as wrong.

Other benchmarks

Benchmark	Drex Max	Drex v2.0	Jev 1.13.0
JevBench (231 public items)	92.2%	85.7%	86.6%
JevBench, hard items	83.8%	73.0%	73.0%
Decision Bench v4 (1,071 items)	94.3%	—	—

License and use restrictions

Drex Max is © Nace.AI, Inc. and is provided under the Nace AI Open RAIL-M License (modified). By using this model you agree to its use-based restrictions (Attachment A), including:

No fully automated decisions without meaningful human review in employment, credit, housing, insurance, healthcare, education, immigration or law enforcement.
No discrimination based on legally protected characteristics.
No medical, legal or financial advice or decisions without qualified human oversight.
No military, weapons, mass-surveillance, fraud, scam or spam use.

Read the full license before use. Drex and Nace are trademarks of Nace.AI, Inc.