{"id":"talos-0-8b","slug":"talos-0-8b","name":"Talos-0.8B","orgId":"tractian","family":"talos-0-8b","releaseDate":null,"availability":"announced","openWeights":true,"license":null,"paramsTotal":800000000,"paramsSource":"slug","contextWindow":null,"knowledgeCutoff":null,"modelCardUrl":"https://huggingface.co/tractian/talos-0.8b","hfRepo":null,"modalities":["text"],"aliases":["Talos-0.8B","talos 0.8b","tractian talos","tractian/talos-0.8b"],"aliasNote":null,"origin":"tracked","tracked":{"trackedSince":"2026-10-22","announced":"2026-10-22","blurb":"Tractian's open decision model for industrial (OT) networks. It sits in the Talos gateway between AI agents, scripts or operator screens and industrial controllers, and scores each typed Modbus or OPC UA request in one forward pass: allow, serve from cache, defer or deny, plus nine other typed questions with calibrated probabilities. It does not generate text. A rank-16 LoRA adapter on Qwen3.5-0.8B-Base, trained on synthetic plant traffic. Weights and code under Apache-2.0.","api":"none","openWeights":"yes","params":{"text":"0.8B (Qwen3.5-0.8B-Base plus a rank-16 LoRA adapter)","source":"Hugging Face model card"},"facts":[{"label":"Base model","value":"Qwen/Qwen3.5-0.8B-Base (Apache-2.0)"},{"label":"Method","value":"Rank-16 LoRA adapter on the base model; answers ten typed questions in one forward pass with calibrated probabilities (model card)"},{"label":"Outputs","value":"action (allow / cache / defer / deny), needed_now, duplicate, servable, over_budget, contradiction, state_conflict, envelope, misbehaving, severity"},{"label":"Protocols","value":"Modbus and OPC UA requests to industrial controllers"},{"label":"Training data","value":"161,602 synthetic labelled requests from simulated Modbus and OPC UA plants (model card)"},{"label":"Out of scope","value":"Not a safety instrumented system and not an authentication mechanism (model card)"},{"label":"Weights","value":"Hugging Face (tractian/talos-0.8b), PEFT adapter format"},{"label":"Code","value":"github.com/tractian/talos (gateway proxy, docs)"},{"label":"License","value":"Apache-2.0"},{"label":"API","value":"None announced"}],"claims":["Lab claim: small enough to run on a network switch, and polices AI-agent behaviour on OT networks.","Lab claim: on a synthetic held-out plant family, 98.2% of bad requests caught, 0.58% of good requests wrongly blocked and 93.6% four-way action accuracy (model card)."],"sources":[{"kind":"repo","title":"Talos-0.8B model card on Hugging Face","url":"https://huggingface.co/tractian/talos-0.8b","date":"2026-10-22"},{"kind":"repo","title":"tractian/talos on GitHub","url":"https://github.com/tractian/talos","date":"2026-10-22"},{"kind":"lab","title":"Announcement on X (Canyon Dell'Omo)","url":"https://x.com/cany0n/status/2113315412314071448","date":"2026-10-22"}],"firstBenchmark":null},"organization":{"id":"tractian","name":"Tractian","slug":"tractian","country":null,"url":"https://tractian.com","color":"#11110F","mark":"T"},"price":null,"answer":{"word":"Pending benchmarks.","scope":"Announced Oct 22, 2026; no independent results yet.","lead":"As of Oct 22, 2026, Talos-0.8B is pending benchmarks on isitsota: announced Oct 22, 2026, with no independent results on the boards we track yet.","confidence":"low"},"status":"PENDING_BENCHMARKS","verdicts":[]}