H Company Releases NeoMME: 260M/800M Multilingual Text-Image Encoders With Day-0 Sentence Transformers Support
lateinteraction · x · 2026-09-07
H Company released NeoMME, multilingual multimodal encoders in 260M and 800M sizes for text and images. The models are day-0 compatible with Sentence Transformers: via MultiVectorEncoder you can search document pages—including charts and tables—without an OCR step. The team says it's looking forward to NeoMME fine-tunes in the wild.
Related event: H Company Open-Sources NeoMME: 260M Multimodal Encoder Matches ColQwen2.5(2 posts)→
More from Multimodal
- NVIDIA's new Sol-H3 fast inference method for H3 awaits a ComfyUI port — krigeta1 · 2026-09-08
- Audio8 Open-Sources On-Device Audio Models From 0.1B to 3B for ASR and TTS — alexcovo_eth · 2026-09-08
- Waypoint 2 Nano enters early access: a world model running locally at 60fps with 50ms latency — cocktailpeanut · 2026-09-08
- Speridlabs releases ENEAS, a text-promptable tracking method claiming gains over SAM3 — joecole · 2026-09-08
- OpenAI demos GPT-6 Astra: from logo generation to interactive design prototypes — OpenAI · 2026-09-08
- Voice Arena Benchmark: Gradium Tops 10 TTS Models With 231ms Median Latency — mattturck · 2026-09-08