{"id":"visionpsy-nano-460m","slug":"visionpsy-nano-460m","name":"VisionPsy-Nano-460M","orgId":"qvac","family":"visionpsy-nano-460m","releaseDate":null,"availability":"ga","openWeights":true,"license":"Apache 2.0 (model cards and LICENSE files); the nanoVLM base is MIT, and the card lists fine-tuning datasets under their own terms, including CC BY-NC-SA 4.0 (TabMWP) and CC BY-NC 4.0 (PKU-SafeRLHF-V)","paramsTotal":507363296,"paramsSource":"tracked","contextWindow":null,"knowledgeCutoff":null,"modelCardUrl":"https://x.com/qvac/status/2082439175891284476","hfRepo":null,"modalities":["image","text"],"aliases":["VisionPsy-Nano-460M","visionpsy-nano","visionpsy nano","visionpsy-nano 460m","visionpsy nano 460m","qvac/visionpsy-nano-460m","visionpsy-nano-460m-flash","visionpsy nano 460m flash","visionpsy-nano flash","qvac/visionpsy-nano-460m-flash","visionpsy-nano-460m-ggufs","visionpsy-nano-460m-flash-ggufs"],"aliasNote":null,"origin":"tracked","tracked":{"trackedSince":"2026-11-02","announced":"2026-07-29","blurb":"A compact vision-language model from Tether's QVAC team for on-device use: it answers questions about a single image (documents, charts, diagrams, scene text) and is post-trained from Hugging Face's nanoVLM-460M-8k, which pairs a SigLIP2 vision encoder with the SmolLM2-360M language model. It comes in two variants, the accuracy-tuned VisionPsy-Nano-460M and VisionPsy-Nano-460M-Flash, which uses far fewer image tokens for faster responses on phones. Weights are on Hugging Face under Apache 2.0, with GGUF builds for llama.cpp; no API is offered.","api":"none","openWeights":"yes","params":{"text":"About 460M per QVAC (the safetensors metadata lists 507,363,296 F32 parameters for each variant)","source":"Hugging Face model card and safetensors metadata for qvac/VisionPsy-Nano-460M and qvac/VisionPsy-Nano-460M-Flash","total":507363296},"facts":[{"label":"Type","value":"Vision-language model: one image plus a text prompt in, text out; English only; multi-image prompts are outside its intended use"},{"label":"Architecture","value":"nanoVLM design: SigLIP2 base patch16-512 vision encoder, pixel-shuffle MLP connector (64 image tokens per 512x512 tile) and the SmolLM2-360M language model; images natively at 512x512, tiled up to 2048 px"},{"label":"Base model","value":"Post-trained from lusxvr/nanoVLM-460M-8k (MIT) in five stages: generic supervised fine-tuning, capability-targeted fine-tuning, weakness-targeted fine-tuning, TIES checkpoint merging and preference alignment with replay"},{"label":"Context","value":"8,192 tokens"},{"label":"Variants","value":"VisionPsy-Nano-460M (tuned for accuracy) and VisionPsy-Nano-460M-Flash (tuned for latency): same architecture and parameter count, separate checkpoints; Flash keeps each image at its native scale before tiling, so a 512x512 image takes 64 image tokens instead of 1,088"},{"label":"Weights","value":"qvac/VisionPsy-Nano-460M and qvac/VisionPsy-Nano-460M-Flash on Hugging Face, F32 safetensors (about 2.0 GB each), loaded in Transformers with trust_remote_code; GGUF builds (FP32 to IQ3, plus a Q8 vision projector) in the -GGUFs repos for llama.cpp"},{"label":"License","value":"Apache 2.0 (model cards and LICENSE files); the nanoVLM base is MIT, and the card lists fine-tuning datasets under their own terms, including CC BY-NC-SA 4.0 (TabMWP) and CC BY-NC 4.0 (PKU-SafeRLHF-V)"},{"label":"Developer","value":"Tether AI Research (Tether Data, S.A. de C.V.), published under the QVAC organization on Hugging Face and X"},{"label":"Evaluation setup","value":"Lab-run VLMEvalKit harness with Qwen3.6-27B as the LLM judge; the VisionPsy support and scoring fixes were open VLMEvalKit pull requests at release"},{"label":"API","value":"None; the weights run locally"}],"claims":["Lab claim: VisionPsy-Nano-460M compares favorably to LFM2.5-VL-450M, SmolVLM2-500M and nanoVLM-460M-8k on 16 of 17 benchmarks (MMVet the clear exception), with a normalized overall score of 62.3 against 59.6, 52.5 and 54.9.","Lab claim: 86.5 on ScienceQA, 42.3 on MM-IFEval and 87.9 F1 on POPE, ahead of the larger FastVLM-0.5B (759M), Qwen3.5-0.8B (873M) and InternVL3.5-1B (1061M) on those three benchmarks.","Lab claim: VisionPsy-Nano-460M-Flash reaches the first token about 19 to 23 times faster than nanoVLM-460M-8k and SmolVLM2-500M on Pixel 9, Galaxy S23 and Galaxy S25 Ultra and up to about 36 times faster on iPhone 15 (4-bit GGUF builds), and 1.3 to 2.4 times faster than LFM2.5-VL-450M.","Lab claim: Flash keeps about 99% of the full model's quality (normalized score 61.4 vs 62.3) with about 40% lower peak memory than nanoVLM-460M and SmolVLM2-500M."],"sources":[{"kind":"lab","title":"QVAC announcement on X","url":"https://x.com/qvac/status/2082439175891284476","date":"2026-07-29"},{"kind":"lab","title":"VisionPsy-Nano (QVAC article on the Hugging Face blog)","url":"https://huggingface.co/blog/qvac/visionpsy","date":"2026-07-29"},{"kind":"lab","title":"QVAC post on X about VisionPsy in a robot demo","url":"https://x.com/qvac/status/2117302716808802705","date":"2026-11-02"},{"kind":"repo","title":"qvac/VisionPsy-Nano-460M on Hugging Face","url":"https://huggingface.co/qvac/VisionPsy-Nano-460M"},{"kind":"repo","title":"qvac/VisionPsy-Nano-460M-Flash on Hugging Face","url":"https://huggingface.co/qvac/VisionPsy-Nano-460M-Flash"},{"kind":"repo","title":"qvac/VisionPsy-Nano-460M-GGUFs on Hugging Face","url":"https://huggingface.co/qvac/VisionPsy-Nano-460M-GGUFs"},{"kind":"repo","title":"qvac/VisionPsy-Nano-460M-Flash-GGUFs on Hugging Face","url":"https://huggingface.co/qvac/VisionPsy-Nano-460M-Flash-GGUFs"},{"kind":"repo","title":"VisionPsy collection (QVAC on Hugging Face)","url":"https://huggingface.co/collections/qvac/visionpsy"}],"firstBenchmark":null},"organization":{"id":"qvac","name":"QVAC (Tether)","slug":"qvac","country":null,"url":"https://qvac.tether.io","color":"#11110F","mark":"Q"},"price":null,"answer":{"word":"Pending benchmarks.","scope":"Announced Jul 29, 2026; no independent results yet.","lead":"As of Nov 22, 2026, VisionPsy-Nano-460M is pending benchmarks on isitsota: announced Jul 29, 2026, with no independent results on the boards we track yet.","confidence":"low"},"status":"PENDING_BENCHMARKS","verdicts":[]}