Open-Weight Models

Model · Time series forecasting

metis-1-preview

Parisi Labs, Inc.

Metis is Parisi Labs' family of world models for physical systems, starting with energy. Metis 1 Preview is its first generation: probabilistic electricity price forecasts that read prices, load, generation and weather and forecast the full range of what power could cost at a location. This card describes the model. Weights are not distributed; forecasts are served through Ask the Grid, the Forecast API and MCP. Price is the target. Recent real-time prices, published day-ahead prices, load, generation and weather explain what the grid is doing, and every input is lined up to when it was actually available. Responses carry the issue time, quantile bands per interval, the settled price once…

YOLO11n-seg weights used by the optional YoloService in deformability-cytometry microscopy application. The.pt checkpoint is the stock Ultralytics YOLO11n-seg release weights (AGPL-3.0). The ONNX export was produced with the bundled script at its default settings (static input shape, no simplification). The application pins this repository by commit and per-file SHA-256 in env/assets.json; scripts/provision-assets.py --asset yolo11n-seg downloads it into the build tree. Bumping the model is a reviewed change to that manifest.

Bubu: ES fine-tune of our own dive1 model75 on the hill (no other entries used) A perpetual policy for the microduck (61-D observation, 14 actions, 50 Hz). Runs until told otherwise — a gait, loaded into a policy slot. The observation normalizer is baked into policy.onnx; feed raw observations. manifest.json follows schema 2 of the microduck policy manifest (docs/policy-manifest.md in the daemon repo).

Mjlab-RollerSprint2m-MicroDuck A perpetual policy for the microduck (61-D observation, 14 actions, 50 Hz, on rollers). Runs until told otherwise — a gait, loaded into a policy slot. The observation normalizer is baked into policy.onnx; feed raw observations. manifest.json follows schema 2 of the microduck policy manifest (docs/policy-manifest.md in the daemon repo). - taskid: Mjlab-RollerSprint2m-MicroDuck Same code, same uv.lock, same command, same seed. Training it again yields a comparable policy, not the same weights: GPU reinforcement learning is not bit-reproducible across machines.

Model · Robotics

microduck-senate_duck_1684

Dennis Schöneberg

2 m Roller Sprint: gladstone fine-tuned by evolution strategies on the Arena's roller scorer. A perpetual policy for the microduck (61-D observation, 14 actions, 50 Hz, on rollers). Runs until told otherwise — a gait for the walk slot. The observation normalizer is baked into policy.onnx; feed raw observations. manifest.json follows schema 2 of the microduck policy manifest (docs/policy-manifest.md in the daemon repo).

Model · Robotics

microduck-sprint-2m-dash1

Stephen E. Tarter

I stopped asking my duck to match a speed and simply paid it for going fast. It learned to throw its heavy head forward and bound down the lane with both legs together, getting a running start before the clock even begins. A perpetual policy for the microduck (61-D observation, 14 actions, 50 Hz). Runs until told otherwise — a gait, loaded into a policy slot. The observation normalizer is baked into policy.onnx; feed raw observations. manifest.json follows schema 2 of the microduck policy manifest (docs/policy-manifest.md in the daemon repo). - taskid: Mjlab-Sprint2m-MicroDuck Same code, same uv.lock, same command, same seed. Training it again yields a comparable policy, not the same…

Model · Robotics

microduck-sprint-es10

Jarrod Pflum

2 m Sprint: suraj-phanindra/microduck-sprint-hc2 (0.892 s, itself from namenu/microduck-arena) fine-tuned by evolution strategies on the Arena sprint scorer (0.888 s). Credit to suraj-phanindra and namenu. A perpetual policy for the microduck (61-D observation, 14 actions, 50 Hz). Runs until told otherwise — a gait, loaded into a policy slot. The observation normalizer is baked into policy.onnx; feed raw observations. manifest.json follows schema 2 of the microduck policy manifest (docs/policy-manifest.md in the daemon repo). A remix of suraj-phanindra/microduck-sprint-hc2.

Model · Robotics

microduck-sprint-hc3

Suraj Phanindra

juicenv/microduck-sprint-es10 fine-tuned by elitist hill climb on the Arena score A perpetual policy for the microduck (61-D observation, 14 actions, 50 Hz). Runs until told otherwise — a gait, loaded into a policy slot. The observation normalizer is baked into policy.onnx; feed raw observations. manifest.json follows schema 2 of the microduck policy manifest (docs/policy-manifest.md in the daemon repo). A remix of juicenv/microduck-sprint-es10.

Model · Robotics

microduck-sprint-v1

Kyrylo Kolesnichenko

sprint-v1 A perpetual policy for the microduck (61-D observation, 14 actions, 50 Hz). Runs until told otherwise — a gait, loaded into a policy slot. The observation normalizer is baked into policy.onnx; feed raw observations. manifest.json follows schema 2 of the microduck policy manifest (docs/policy-manifest.md in the daemon repo). A remix of HannesVonEssen/microduck-running.

VibeVoice-Realtime is a lightweight real‑time text-to-speech model supporting streaming text input and robust long-form speech generation. It can be used to build realtime TTS services, narrate live data streams, and let different LLMs start speaking from their very first tokens (plug in your preferred model) long before a full answer is generated. It produces initial audible speech in ~300 ms (hardware dependent). ▶ Watch demo video (Launch your own realtime demo via the websocket example in Usage) Although the model is primarily built for English, we found that it still exhibits a certain level of multilingual capability—and even performs reasonably well in some languages. We provide nine…

The base model pretrained on 16kHz sampled speech audio. When using the model, make sure that your speech input is also sampled at 16kHz. Note: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model speech recognition, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out this blog for more in-detail explanation of how to fine-tune the model. The model was pre-trained on 960h of Librispeech. Authors: Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng…

Model

minecraft-world-jepa-ckpts

Keylhan Paumard--André

Autoregressive video world model for Minecraft. A frozen Stable Diffusion VAE compresses 256×256 frames to 4×32×32 latents; a decoder-only transformer (338M params, RoPE, DWS-style per-block action adapters, SDPA flash attention) predicts the next latent conditioned on discrete action tokens (29 ids: 22 keyboard + 6 camera bins + 1 no-op). Each run directory holds a full resumable state (model + optimizer + scheduler + step/epoch counters), so any file below restarts training bit-identically with the matching config and --resume: Weights derive from public TESS-VLA gameplay data and an sd-vae-ft-mse base. No license is asserted here — check the dataset and base-model terms before commercial…

Ming（AI 算命 App）端侧推理用的 MNN 模型包，zip 打包，一次取回即可用。 zip 内是扁平结构，解压即得模型目录内容： - qwen3-0.6b：llm.mnn · qwen3.5-0.8b：上面 4 （该仓库的 llmconfig.json 声明 isvisual: true，缺视觉权重会加载失败） 每个文件的 sha256 与 App 随包配置 assets/models.json 中的声明一致（打包前逐个校验通过）。