Meet Drex Max: Nace AI’s Multimodal Decision Model for Computer-Use Agents with a Score 67.59

Nace AI has released Drex Max, a multimodal decision model for agents that use computers and browsers. The Drex Max multimodal decision model reads text and images and returns calibrated probabilities. It writes no text. Nace team reports 67.59 on the public Decision Index 0.3.2, 6.9 points above TypeSafe’s Jev 1.14. For agent builders, that means one cheap, typed call per click.

TL;DR

Size: Parameters not disclosed. Up to 262,144 tokens per question on the API, including image tokens.
Runs on: Hosted API only (drex-max on Nace Console). No open weights or quantized builds announced.
Performance: 67.59 on Decision Index 0.3.2 (public), the top score Nace lists for any decision model.
Best: 98.2 on BFCL tool calling, and 92.2 on JevBench versus 86.7 for Jev 1.14.
Worst: 19.0 on HLE, and 72.5 on MMLU-Pro versus 82.1 for Jev 1.14.

What is Drex Max?

Drex Max is a decision model from Nace AI that answers typed questions about text and images. It is the most accurate model in the Drex family, and the only one that reads images.

A decision model does not generate tokens. You send a state and named questions. Each question is one of 3 types: noul (yes/no), choice (pick a label) or score (ordered scale). The model returns a probability distribution for each answer in 1 HTTP call. Drex can only answer with the options you gave it.

How does Drex Max work with images?

Drex Max uses the same POST /v1/systemone endpoint as every Drex model. You select it with “model”: “drex-max”. It is a pinned ID, so drex-latest does not route to it.

Images go in an images array. Every question sees the same images, before the state and questions. You refer to them as ‘Image 1’ and ‘Image 2’ in plain text. Each entry is either an uploaded workspace file or inline base64.

Key image rules from the Drex Max guide:

Formats: still PNG, JPEG or WebP.
Up to 8 images per request, 16 MiB each, 32 MiB total.
Each image is resized to at most 1,638,400 pixels (1,280 × 1,280).
Each image costs up to 1,602 tokens in every question.
Up to 16 questions per request, with a 4 MiB request body.

How does Drex Max perform on benchmarks?

All scores below come from Nace’s Drex page. They are self-reported. Nace says Drex trains on the index’s official training splits and is scored on held-out splits.

Decision benchmarks: Drex Max scores 67.6 on the public Decision Index 0.3.2. Jev 1.14 scores 60.7 and Drex 2.0 scores 61.3. On JevBench, Drex Max reaches 92.2 against 86.7 for Jev 1.14.
Agentic benchmarks: Drex Max posts 0.780 on OSWorld-Verified and 98.2 on BFCL. It scores 85.2 on When2Call and 84.8 on RAGTruth.
Vision: On the Decision Index Vision public suite, Drex Max scores 74.5. GPT-6 Luna Decisions scores 63.9. Mind2Web reaches 88.6 and InfographicVQA reaches 98.8.
Where it trails: Jev 1.14 leads on MMLU-Pro (82.1 vs 72.5), HLE (19.4 vs 19.0) and ARC-Challenge (97.6 vs 97.4). On Hateful Memes, Drex Max trails GPT-6 Luna Decisions, 70.6 vs 71.2.

<div class="tabs" role="tablist">
<button class="tab" role="tab" aria-selected="true" data-t="flow">How it works</button>
<button class="tab" role="tab" aria-selected="false" data-t="bench">Benchmarks</button>
<button class="tab" role="tab" aria-selected="false" data-t="vs">vs Jev 1.14</button>
<button class="tab" role="tab" aria-selected="false" data-t="lim">API limits</button>
<button class="tab" role="tab" aria-selected="false" data-t="cost">Cost calculator</button>
</div>

<div class="pn on" data-p="flow">
<div class="flow">
<div class="node" data-n="0"><h4>1 · Input</h4>
<p><code>state</code>: text or JSON</p>
<p><code>images</code>: up to 8, still PNG / JPEG / WebP</p>
<div class="imgs"><span></span><span></span><span></span></div>
</div>
<div class="arr"><i></i></div>
<div class="node" data-n="1"><h4>2 · Questions</h4>
<p><code>noul</code> yes / no</p>
<p><code>choice</code> pick a label</p>
<p><code>score</code> ordered scale</p>
<p style="color:#a8a29e">Up to 16 per request</p>
</div>
<div class="arr"><i></i></div>
<div class="node" data-n="2"><h4>3 · Answer</h4>
<p style="margin-bottom:4px">"Which button next?"</p>
<div class="prob"><em>search</em><div class="b"><u data-w="81"></u></div><b>0.81</b></div>
<div class="prob"><em>cart</em><div class="b"><u data-w="12"></u></div><b>0.12</b></div>
<div class="prob"><em>back</em><div class="b"><u data-w="7"></u></div><b>0.07</b></div>
</div>
</div>
<p class="cap" id="mtp-cap"></p>
<div class="ctl">
<button class="btn" id="mtp-prev"> Back</button>
<button class="btn pri" id="mtp-next">Next step </button>
<button class="btn" id="mtp-play">Auto play</button>
<div class="dots"><i></i><i></i><i></i></div>
</div>
<p class="note">Answer probabilities are an illustration of the response shape, not real model output.</p>
</div>

<div class="pn" data-p="bench">
<div class="seg" id="mtp-seg">
<button class="on" data-s="di">Decision Index</button>
<button data-s="ag">Agentic</button>
<button data-s="vi">Vision</button>
</div>
<div id="mtp-bars"></div>
<p class="note" id="mtp-bnote"></p>
</div>

<div class="pn" data-p="vs">
<table class="vs">
<thead><tr><th>Benchmark</th><th style="text-align:right">Drex Max</th><th style="text-align:right">Jev 1.14</th><th>Result</th></tr></thead>
<tbody id="mtp-vs"></tbody>
</table>
<p class="note">All figures from Nace’s Drex page (self-reported). Jev 1.14 rejects 15.9% of screenshots by size, per Nace.</p>
</div>

<div class="pn" data-p="cost">
<div class="calc">
<div>
<label for="mtp-dec">Decisions per day: <b id="mtp-decv" style="color:#f5f5f4">10,000</b></label>
<input type="range" id="mtp-dec" min="1000" max="1000000" step="1000" value="10000">
<label for="mtp-img">Screenshots per decision: <b id="mtp-imgv" style="color:#f5f5f4">1</b></label>
<input type="range" id="mtp-img" min="0" max="8" step="1" value="1">
<label for="mtp-txt">Text tokens per decision (state + question): <b id="mtp-txtv" style="color:#f5f5f4">500</b></label>
<input type="range" id="mtp-txt" min="50" max="20000" step="50" value="500">
</div>
<div class="out">
<small>Estimated daily cost, upper bound</small>
<div class="big" id="mtp-cost">$0.00</div>
<small id="mtp-tok"></small>
<small style="margin-top:10px">Uses $0.06 per 1M input tokens and the 1,602-token maximum per image. Output tokens are not billed. Real image token counts can be lower.</small>
</div>
</div>
</div>

(function(){var f=document.getElementById(‘mtp-drexmax-frame’);window.addEventListener(‘message’,function(e){if(!e.data||e.source!==f.contentWindow)return;var h=e.data.mtpDrexMaxHeight;if(h&&h>200)f.style.height=h+’px’;});})();

How fast and how cheap is Drex Max?

Nace lists a 174 ms median per screenshot decision. Jev 1.14 takes 253 ms on the same 243 questions. Nace says Drex Max accepts every screenshot size, while Jev 1.14 rejects 15.9% by size.

Across its benchmark table, Nace reports a 183 ms median latency. The launch post cites 34 ms per decision and under 1 second at 262K tokens. Those figures use different setups, so treat them as separate claims.

Pricing is $0.06 per 1M input tokens. Output tokens are not billed. New accounts get $25 in signup credit, and failed requests cost nothing.

How does Drex Max compare with other decision models?

Note: d3-flash’s vision figure is its own evaluation on the 0.3.2 vision board. Drex Max’s figure is Nace’s own run. Cross-vendor numbers are not a controlled comparison.

Key Takeaways

Drex Max is Nace AI’s first multimodal decision model.
It scores 67.59 on Decision Index 0.3.2, 6.9 above Jev 1.14.
Screenshot decisions take 174 ms, versus 253 ms for Jev 1.14.
Pricing is $0.06 per 1M input tokens, with output free.
Size is undisclosed, weights are closed, and scores are self-reported.

Source link

What's Hot

Robinhood links 29 million app users to Ripple’s XRP Ledger

Tsinghua AI Open-Sources AKV, a Cache Engineering Server for LLM Agents on SGLang

ENS DAO Approves ENSv2 With 1.69M Votes, Execution Due October 27

Meet Drex Max: Nace AI’s Multimodal Decision Model for Computer-Use Agents with a Score 67.59

Tsinghua AI Open-Sources AKV, a Cache Engineering Server for LLM Agents on SGLang

Claude Bypassed Georgia’s Paywall to Pull Vote Count Data

At MIT, Martha Minow speaks on sustaining democracy, both by using and overcoming AI | MIT News

Robinhood links 29 million app users to Ripple’s XRP Ledger

ENS DAO Approves ENSv2 With 1.69M Votes, Execution Due October 27

Bitcoin ETFs Face Roughest Run Since April as Banks Stay Shut

Letest News

Robinhood links 29 million app users to Ripple’s XRP Ledger

Tsinghua AI Open-Sources AKV, a Cache Engineering Server for LLM Agents on SGLang

What's Hot

Meet Drex Max: Nace AI’s Multimodal Decision Model for Computer-Use Agents with a Score 67.59

TL;DR

What is Drex Max?

How does Drex Max work with images?

Key image rules from the Drex Max guide:

How does Drex Max perform on benchmarks?

How fast and how cheap is Drex Max?

How does Drex Max compare with other decision models?

Key Takeaways

Related Posts