Meta releases MetaEncoder: unified text-image-video encoder with SOTA retrieval
algo_diver · x · 2026-11-13
Meta released MetaEncoder on Hugging Face, a unified multimodal encoder for text, images, and video targeting retrieval and decision support, claiming SOTA results.
Benchmark table highlights (MetaEncoder v1):
- Leads on MMLU-Pro (0.671), MMMU (0.665), MMMU-Pro (0.513), VideoMMMU (0.646), NaturalBench (0.832), and TempCompass (0.778) among compared models;
- Tops open-set retrieval with NanoBEIR 0.701, MMEB V3 image 0.816, and video 0.743, only trailing WeMM-Embedding (9B) on visdoc retrieval;
- Compared against OpenJev, Jev-Omni, WeMM-Embedding (9B), and Qwen3-VL-Embedding (8B); OpenJev still leads on JevBench hard (0.766).
Related event: Meta Open-Sources MetaEncoder, a Unified Multimodal Encoder for Retrieval(3 posts)→
More from Models
- Raon-1 nano tops consumer-GPU agent tests: 96/96 tasks, 187 tok/s at ~2GB Q4 — Kangwook_Lee · 2026-11-13
- Same harness, same rules: Argon used the code graph in 12/17 sessions, Opus only 4/17 — rseroter · 2026-11-13
- Massive eval table update benchmarks open-weight models across coding, finance, legal — siyan_zhao · 2026-11-13
- Did Kimi K3 and DeepSeek-V4-Pro each train on 2,000-3,000 Blackwells? Claim disputed — teortaxesTex · 2026-11-13
- Community notes: Ling's Blackwell usage was an open secret, papers admitted renting compute — teortaxesTex · 2026-11-13
- Gemini 4 Hydrogen hits frontier-level scores at 1/3 the cost of Anthropic's flagship — Electronic-Weekend66 · 2026-11-13