Tencent's Hy-Nomas weights land on Hugging Face under Apache 2.0 with 280-case benchmark
lmoroney · x · 2026-10-25
Follow-up on the Hy-Nomas open-source release: weights are on Hugging Face (tencent/Hy-Nomas) under Apache 2.0 with Diffusers/Safetensors/PyTorch formats. Architecture: segment-autoregressive DiT + Qwen text encoder + camera conditioning + PixelUMM decoder, distilled to 8 denoising steps at 720×1280/24FPS, plus a single-RTX-4090 config (INT8 DiT weights and KV cache, FP8 decoder, CPU offload). Also ships Hy-Nomas-Bench: four benchmarks with 280 test cases.
More from Multimodal
- Segmenting every corn kernel with Muse Wonder + SAM 3.1 on Meta's Model API — keerthanpg · 2026-10-26
- Dense instance segmentation via Muse Wonder + SAM 3.1 on Meta's Model API — nikhilaravi · 2026-10-26
- LichtFeld Studio v0.5.4.3 ships community-built SteadySplats, POPSpa and Voxelization — janusch_patas · 2026-10-26
- One prompt to a 2-3 min narrated short: hands-on test of Antigravity video generation — AI_Andrew · 2026-10-26
- Creator makes 10,564-frame animated short entirely with Gemini toolchain — AI_Andrew · 2026-10-26
- FLUX.3 Klein step-distilled to 4-8 steps with CFG baked in — Fit-Structure-6489 · 2026-10-26