--- title: "Towards Foundation Decision Models: Drex" description: "Our technical report on Drex, four open decision models that return a calibrated probability for every option instead of generating text. Drex Max sets a new state of the art on Decision Index 0.4.1 and is competitive with frontier models on real-data decisions in under 200 ms." canonical: "https://nace.ai/research/towards-foundation-decision-models" updated: 2026-11-06 --- # Towards Foundation Decision Models: Drex > Our technical report on Drex, four open decision models that return a calibrated probability for every option instead of generating text. Drex Max sets a new state of the art on Decision Index 0.4.1 and is competitive with frontier models on real-data decisions in under 200 ms. Research · Nov 6, 2026 · Nischay Dhankhar, Alisher Amantay, Zhanibek Datbayev, Darkhan Akhmetov, Yash Shah, Nurmukhamed Ubaidullayev, Raman Saparkhan, Dos Baha, Abulhair Saparov # Towards Foundation Decision Models: Drex A lot of what AI does in production isn't writing. It's choosing: which team gets this ticket, which button the agent clicks next, whether this passage answers the question. The usual way to do that is to ask a large model, let it think, and parse an answer out of its text. That's slow, costly, and gives different answers to the same input. A **decision model** skips the text. You pass a context, a question, and your options, and one forward pass returns a probability for every option. It can't answer outside the set you gave it. Today we're publishing the technical report on **Drex**, our family of four open decision models, and on how we built the largest of them, **Drex Max**. It is one of the first in-depth technical reports on how decision models behave. Here are a few results we're most excited about. ![Bar chart of Decision Index 0.4.1 scores: Drex Max 67.52 public and 66.60 full, Drex v2.0 61.25 and 52.76, Drex v1.5 58.08 and 45.78, Drex DLM 52.31, Jev 1.14.0 60.70 and 60.24, GPT-6 Luna Decisions 47.78.](https://nace.ai/images/research/towards-foundation-decision-models/decision-index.webp) *Decision Index 0.4.1, public suite and full score with private tests. Drex models in blue.* **A new state of the art, and it holds on unseen tests.** Drex Max leads the decision models we compare on Decision Index 0.4.1, with 67.52 on the public suite against 60.70 for Jev 1.14.0. When the private tests are added, its score drops less than a point, demonstrating strong generalization capabilities. ![Grouped bar chart of chance-corrected skill by domain for Drex Max, Drex v2.0, Drex v1.5, Jev 1.14.0, and GPT-6 Luna Decisions. Drex Max scores 57 in Knowledge, 72 in Language, 69 in Retrieval, 82 in Tools, and 53 in Arts.](https://nace.ai/images/research/towards-foundation-decision-models/domain-skill.webp) *Per-domain skill on Decision Index 0.4.1, public suite. Drex Max leads in every domain except Arts, where v2.0 is 0.5 points higher, and Knowledge, where it ties Jev 1.14.0.* **Independent results agree.** On the Artificial Analysis Decision Index, Drex Max scores 64, within two points of the three leading decision models. **Frontier accuracy, without the wait.** On Decision Bench v4, with raw image input, Drex Max is statistically tied with d3-turbo and GPT-6 Sol and 1.6 points behind Claude Opus 5.5. It answers a typical request in 183 ms. GPT-6 Sol takes about 2 seconds and Opus about 3.5. **It holds up on other decision benchmarks too.** Drex Max is the most accurate decision model on DecideBench, close to the top decision models on JevBench-231, and the only model that clearly beats a transcript-length heuristic when judging OSWorld runs. ![Four dot charts. JevBench-231 accuracy: Claude Opus 5.5 100.0, GPT-6 Sol 99.1, Drex Max 91.8. DecideBench pair accuracy: GPT-6 Sol and Claude Opus 5.5 98.5, Drex Max 97.0, Jev 1.14 96.0. Game Bench win rate: GPT-6 Sol 65.0, Drex Max 64.0, Claude Opus 5.5 63.0. OSWorld-Verified AUROC: Drex Max 0.780, Jev 1.13.0 0.734.](https://nace.ai/images/research/towards-foundation-decision-models/other-benchmarks.webp) *Other decision benchmarks. Drex models in purple; axes don't start at zero. The dashed lines mark the Jev 1.14 reference in the game arena and the transcript-length heuristic on OSWorld-Verified.* **Calibrated confidence.** Drex models return a probability for every option, calibrated so that confidence tracks how often the model is right. When Drex Max is at least 93% sure, it is right 97% of the time. So you can rely on that confidence when building workflows and pipelines. ![Calibration curves of accuracy against predicted probability for Drex Max, Drex v2.0, Drex v1.5, and GPT-6 Luna Decisions. All four stay close to the diagonal.](https://nace.ai/images/research/towards-foundation-decision-models/calibration.webp) *Calibration curves on Decision Index 0.4.1. A perfectly calibrated model lies on the dashed diagonal.* **Answer consistency.** When the answer options in a prompt are randomly reordered, only 4.3% of Max's answers change, and accuracy barely moves. The answers that do flip are mostly ones the model already flagged as unsure. **It reads the image, not a description of it.** On image-only questions, Max is right 63% of the time from a text description of the image and 98% from the raw image. ## In the paper The report covers how every Drex model is built. The three pointer models encode the input once and score every question's options with a small pointer head. Drex Max is a 27B multimodal model with bidirectional attention and a linear readout head over the option codes. ![Diagram of the pointer models, which score options with a pointer head over a cached state, and Drex Max, which reads images and lettered options in one sequence and answers with a readout head.](https://nace.ai/images/research/towards-foundation-decision-models/architecture.webp) *Pointer models (left) share one cached state across questions. Drex Max (right) reads each question in full and answers with a readout head.* The report also covers the training recipe, including RLAF, which rewards verified answers adjusted by a judge's probabilities without drifting from the supervised model, and adds 2 index points over supervised training alone. It closes with full evaluations against decision-specialized and frontier models, how we serve the models in production, and the limitations we still see, closed-book knowledge first among them. All four models are open source. Weights and inference code are on [GitHub](https://github.com/nace-ai/drex-decision-models) and [Hugging Face](https://huggingface.co/nace-ai), and Drex Max is served through the [Drex API](https://nace.ai/drex). ## For agents - [Drex API guide](https://nace.ai/drex.md): the decision model: request shape, question types, benchmarks and deployment, as markdown - [Decision endpoint](https://console.nace.ai/v1/systemone): POST with model "drex-latest", a state string and typed questions (choice, noul, score); https://console.nace.ai/v1/models lists checkpoints - [Document endpoints](https://console.nace.ai/v1/documents/{parse|split|classify|extract|ground}): POST a source (a public URL, an uploaded workspace file or a finished parse job) and poll GET https://console.nace.ai/v1/documents/jobs/{id}; the same key and credit as decisions - [Console](https://console.nace.ai): API keys, the playground, usage and billing; send `Authorization: Bearer $NACE_API_KEY` on every request - [Console docs](https://console.nace.ai/docs): quickstart, question types, the document routes, errors, limits and pricing - [Nace Document Intelligence](https://nace.ai/perception.md): the Perception series: parse, split, classify, extract and ground any file type with a location on every value - [Drex for Search](https://nace.ai/drex/search.md): Drex for Search: decision-powered retrieval for agents, Drex finds the page, cited search through the CLI, SDKs, MCP and REST, and the public benchmark ranks - [Use cases](https://nace.ai/use-cases.md): every Drex use case with its proof figures: browser use, search and finance for developers, enterprise finance teams and professional firms - [Developers](https://nace.ai/developers.md): The execution layer for agents: Perception, Decision and Search behind one key, the Drex Max frontier, install steps, a scripted agent session and the four deployment boundaries - [Site index](https://nace.ai/llms.txt): every page on nace.ai; append `.md` to any page URL or send `Accept: text/markdown` for the markdown version