{"id":"sophon-realtime","slug":"sophon-realtime","name":"Sophon-Realtime v1","orgId":"adventists-ai","family":"sophon-realtime","releaseDate":null,"availability":"ga","openWeights":true,"license":"Apache-2.0 for Sophon-Realtime code and weights (GitHub LICENSE, HF cards, PyPI)","paramsTotal":32000000000,"paramsSource":"slug","contextWindow":null,"knowledgeCutoff":null,"modelCardUrl":"https://github.com/adventists-ai/sophon-realtime","hfRepo":null,"modalities":["text","audio","image"],"aliases":["Sophon-Realtime v1","sophon-realtime","Sophon-Realtime","sophon-realtime-v1","Sophon-Realtime-32B-Tag","Sophon-Realtime-32B-Flash","Sophon-Realtime-32B-E2E","adventists-ai/Sophon-Realtime-32B-Tag","adventists-ai/Sophon-Realtime-32B-Flash","AdventistsAI/Sophon-Realtime-32B-Tag","pip:sophon-realtime"],"aliasNote":null,"origin":"tracked","tracked":{"trackedSince":"2026-11-09","announced":"2026-11-08","blurb":"Adventists AI's open full-duplex speech-to-speech family. Tag, Flash and an E2E research preview share one 32B thinker that listens continuously, decides turn-taking and writes the reply; they differ in how the reply becomes sound. Apache-2.0 code and weights are public on Hugging Face and ModelScope for self-hosting with vLLM-Omni (Tag/Flash on one 80 GB GPU) or the lab's E2E engine. A ModelScope Studio demo exists; no separate public hosted API is confirmed.","api":"none","openWeights":"yes","params":{"text":"32B thinker (shared); Tag/Flash/E2E speaking stacks differ","source":"Adventists AI README and Tag model card"},"facts":[{"label":"Lab","value":"Adventists AI (GitHub/HF org adventists-ai; ModelScope AdventistsAI). Contributor accounts include BKtensor / @BKJinAI"},{"label":"Role","value":"Open full-duplex voice model: continuous listening, barge-in, back-channels, and spoken replies with optional non-verbal sound tags"},{"label":"Builds","value":"Tag (recommended expressive speech model; ~0.6 s first audio), Flash (lighter/faster speech model; 0.31 s first audio), E2E preview (talker inside the 32B model; ~1 s; two 80 GB GPUs)"},{"label":"Lineage","value":"Thinker is DuplexJev-32B-Para v1.5 with tag writing merged in (FP8, vision tower retained). DuplexJev's own card lists Qwen3-VL-32B-Instruct language-model weights plus a Qwen3-ASR-0.6B encoder — not a plain Qwen3-32B text checkpoint"},{"label":"License","value":"Apache-2.0 for Sophon-Realtime code and weights (GitHub LICENSE, HF cards, PyPI). The standalone DuplexJev-32B-Para model card is still tagged CC-BY-NC-4.0 — licensing discrepancy versus this Sophon redistribution"},{"label":"Access","value":"Public ungated HF/ModelScope weights (Tag/Flash/E2E); GitHub release v1.0; pip package sophon-realtime (Tag serve CLI). ModelScope Studio demo linked from the README. Self-host only for production serving; no separate public hosted API confirmed"},{"label":"Serving","value":"Tag/Flash: one 80 GB GPU via vLLM-Omni (lab recommends up to 16 / 40 concurrent users). E2E: own engine on two 80 GB GPUs. Do not run third-party installers from this tracking workflow"},{"label":"APIs when self-hosted","value":"ws://HOST:PORT/v1/realtime (OpenAI-Realtime-style full duplex) and POST /v1/chat/completions (audio in, text+audio out); optional Live mode compatible with OpenAI GPT-Live clients"},{"label":"Limitations (lab)","value":"Tag card: [sigh] and laughter spans remain weak versus how often the thinker writes them; Mandarin slightly less clear than English with the distilled flow; very long replies (over about a minute of speech in one turn) exceed voice training"},{"label":"Related","value":"Family page covers Tag/Flash/E2E. Keep separate from DuplexJev hearing models, tag-tts, and Qwen3-VL-32B-Instruct"},{"label":"Date evidence","value":"GitHub release v1.0 published 2026-11-08T05:48:49Z (12:48 AM ET) with Tag/Flash weights linked; HF Tag/Flash family card commits the same morning. Oct 29 HF createdAt and Nov 7 weight-upload commits are not used without proof the repo was public then; the repo's first commit was a teaser saying it opens Nov 9"}],"claims":["Lab claim: self-reported VoiceBench 82.8 and Big Bench Audio 81.8% for the shared 32B brain (README / family card; self-run unless noted).","Lab claim: Flash first audio 0.31 s; Tag about 0.6 s (README table; one-user figures).","Lab claim: Tag NVV-SuperBench 0.86 Chinese / 0.80 English; Full-Duplex-Bench v3 tool-calling Pass@1 48% with 100% response rate (README; self-run)."],"sources":[{"kind":"repo","title":"Adventists AI Sophon-Realtime GitHub (Apache-2.0)","url":"https://github.com/adventists-ai/sophon-realtime"},{"kind":"repo","title":"Sophon-Realtime v1.0 GitHub release","url":"https://github.com/adventists-ai/sophon-realtime/releases/tag/v1.0","date":"2026-11-08"},{"kind":"repo","title":"Sophon-Realtime-32B-Tag model card and weights","url":"https://huggingface.co/adventists-ai/Sophon-Realtime-32B-Tag"},{"kind":"repo","title":"Sophon-Realtime-32B-Flash model card and weights","url":"https://huggingface.co/adventists-ai/Sophon-Realtime-32B-Flash"},{"kind":"repo","title":"Sophon-Realtime v1 family card","url":"https://huggingface.co/adventists-ai/Sophon-Realtime-v1"},{"kind":"repo","title":"DuplexJev-32B-Para card (lineage and CC-BY-NC-4.0 tag)","url":"https://huggingface.co/adventists-ai/DuplexJev-32B-Para"},{"kind":"lab","title":"Author announcement on X","url":"https://x.com/BKJinAI/status/2119673787700699235","date":"2026-11-09"}],"firstBenchmark":null},"organization":{"id":"adventists-ai","name":"Adventists AI","slug":"adventists-ai","country":null,"url":"https://github.com/adventists-ai/sophon-realtime","color":"#11110F","mark":"A"},"price":null,"answer":{"word":"Pending benchmarks.","scope":"Announced Nov 8, 2026; no independent results yet.","lead":"As of Nov 22, 2026, Sophon-Realtime v1 is pending benchmarks on isitsota: announced Nov 8, 2026, with no independent results on the boards we track yet.","confidence":"low"},"status":"PENDING_BENCHMARKS","verdicts":[]}