v4: +56 multimodal trending (FLUX.1/2, SD-3.5, Z-Image, ERNIE-Image, HunyuanImage/Video, LTX-2.3, Wan2.2, VoxCPM2, Kokoro-82M, Voxtral-TTS, Qwen3-TTS, whisper-v3-turbo, granite-speech, ACE-Step, Gemma-4-E, Qwen3-Omni, MiniCPM-o-4_5, Phi-4-mm, etc.)

README.md

@@ -15,23 +15,32 @@ short_description: Darwin family + 2026 trending models on the HF galaxy
 Interactive **model ontology / lineage galaxy** built on top of [Eliahu/Model-Atlas](https://huggingface.co/spaces/Eliahu/Model-Atlas) (arXiv 2503.10633), with two custom overlays:
 
 1. **🌳 VIDRAFT Darwin family** (120 models + 41 anchors + 120 edges) — the entire Darwin / Rogue / AWAXIS / TenOS / Warecube / AETHER / NOESIS-Darwin ecosystem, with parent → child merge / quantize / adapter chains
-2. **⭐ 2026 Trending originals** (48 models + 17 anchors + 19 edges) — DeepSeek-V4 family, MiMo-V2.5, Hy3-preview, GLM-5.1, Hermes-4.3-36B, Kimi-K2, gpt-oss family, NVIDIA Nemotron-3 (Super-120B / Nano-30B / Omni / Cascade), Qwen3-Coder-Next, Llama-3.3-70B, Granite-4.1-8B, VibeVoice, dots.mocr, CADD-Base-7B, etc.
+2. **⭐ 2026 Trending NLP** (48 models, lower-right) — DeepSeek-V4 family, MiMo-V2.5, Hy3-preview, GLM-5.1, Hermes-4.3-36B, Kimi-K2, gpt-oss family, NVIDIA Nemotron-3 (Super-120B / Nano-30B / Omni / Cascade), Qwen3-Coder-Next, Llama-3.3-70B, Granite-4.1-8B, dots.mocr, CADD-Base-7B, etc.
+3. **🎨 2026 Trending Multimodal** (56 models, left wedge cluster) — by category:
+   - **Image gen**: FLUX.1-dev/schnell, ERNIE-Image (+Turbo), Z-Image (+Turbo), SD-3.5-large, SDXL-base, PixelDiT-1300M
+   - **Image edit**: FLUX.2-dev/klein-4B/9B, FLUX.1-Kontext, Qwen-Image-Edit-2509/2511, HunyuanImage-3.0
+   - **Video**: LTX-2 / LTX-2.3, Wan2.2-TI2V-5B / I2V-A14B, HunyuanVideo-1.5, LongCat-Video, Motif-Video-2B, SVD-img2vid
+   - **TTS+LLM**: OmniVoice (k2-fsa), VoxCPM2, Kokoro-82M, Voxtral-4B-TTS, sarashina2.2-tts, fishaudio s2-pro, XTTS-v2, MOSS-TTS-Nano, VibeVoice-Realtime, CosyVoice3, Qwen3-TTS-12Hz (1.7B/0.6B), chatterbox
+   - **Speech/ASR**: VibeVoice-ASR, granite-speech-4.1 (2b/+plus), whisper-large-v3 (+turbo), cohere-transcribe-03-2026, Qwen3-ASR-1.7B, parakeet-tdt-0.6b-v3
+   - **Audio gen**: ACE-Step1.5 / acestep-v15-xl-turbo
+   - **Any-to-Any (Omni)**: LLaDA2.0-Uni, SenseNova-U1-8B-MoT, Gemma-4-E4B/E2B-it, MiniCPM-o-4_5, Qwen3-Omni-30B-A3B, HKUSTAudio Audio-Omni, Phi-4-multimodal-instruct
 
 ## Atlases
 
 | Atlas | Nodes | Edges | Purpose |
 |---|---:|---:|---|
-| **Modalities Atlas (+ Darwin + Trending)** *(default)* | 22 427 | 24 225 | Full HF modality landscape with both overlays |
-| Large NLP Atlas (+ Darwin + Trending) | 137 271 | 152 917 | Full HF NLP ecosystem with both overlays |
-| 🌳 DARWIN Family Only | 223 | 139 | Compact Darwin ontology + connected trending models |
+| **Modalities Atlas (+ Darwin + Trending + Multimodal)** *(default)* | 22 482 | 24 234 | Full HF modality landscape with all overlays |
+| Large NLP Atlas (+ Darwin + Trending + Multimodal) | 137 326 | 152 926 | Full HF NLP ecosystem with all overlays |
+| 🌳 DARWIN Family + Trending compact view | 279 | 148 | Compact custom-injected ontology |
 | Hugging Face Galaxy | full | full | Original Eliahu galaxy view (untouched) |
 
 ## Cluster placement
 
 | Cluster | Coordinates | Size |
 |---|---|---:|
-| 🌳 Darwin (upper-right) | x ≈ 7000, y ≈ 5000 | 120 nodes |
-| ⭐ 2026 Trending (lower-right) | x ≈ 6500, y ≈ -5500 | 48 nodes |
+| 🌳 Darwin family (upper-right) | x ≈ 7000, y ≈ 5000 | 120 nodes |
+| ⭐ 2026 Trending NLP (lower-right) | x ≈ 6500, y ≈ -5500 | 48 nodes |
+| 🎨 2026 Trending Multimodal (left wedges) | x ≈ -6500, y ≈ -1000 | 56 nodes |
 
 Both clusters are far enough from the existing HF nodes that they are immediately spotted on first load.
 

compressed_darwin_data.json.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c4f860e1530b981e566b0cd5a787316f79d246e97a27d0f8b80837acc862a38b
-size 25244
+oid sha256:4a28a712a94e087c38df0c51280e6ef2986ab6b2298552acf913f2b2a91eee9f
+size 32625

compressed_large_nlp_data.json.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2bbe96f224315b9fc84bd8a0135df70506bb0061d58c6422578d3f3f1f69e92b
-size 23670602
+oid sha256:915b9294297d5574a1c13e32967122f3dbc5b11837b75245682bb21d021a6111
+size 23677749

compressed_modalities_data.json.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cdf5038ade68b9e4dcfd2b0ba15d2e04ff85ecdf76eccf9eca5f352b8ccddca0
-size 3121171
+oid sha256:e6220bf2e676ce9a98196be990ce8a9764bafa41a8846c17b88a660fe342821d
+size 3128145