WeChat unveils WeLM-Image: unified 31.7B-param image model tops open-source benchmarks
Canmiao Fu · hf · 2026-11-06
WeChat Vision (Tencent) releases the technical report for WeLM-Image, a unified image foundation model integrating multimodal understanding into text-to-image generation, instruction-guided editing, and multi-reference synthesis:
- Mixture-of-Transformers (MoT) backbone with shared attention and separate understanding/generation parameters; per-path MoE modules give 31.7B stored params and 5.4B activated per target latent token.
- A generation-oriented data engine pairs a knowledge-grounded visual vocabulary with large-scale visual clustering; bilingual annotations support progressive training up to native 1.5K resolution.
- Generation and editing are optimized separately with rubric/checklist rewards, then merged via noise-aware on-policy distillation.
- The 1536-resolution model ranks first among evaluated open-source models on PRISM-Bench, Qwen-Image-Bench, and UnicBench.
- MeanFlow few-step distillation yields WeLM-Image-Turbo: 1024×1024 images in 4.98s on a single NVIDIA H20 with 8 guidance-free steps, designed for single-GPU deployment in WeChat's consumer apps.
More from Multimodal
- AI-generated UGC-style ad creatives shown off on Reddit — Scope_Cycle · 2026-11-06
- Ant's Ling-3.2-flash: 560B-param native multimodal model beats Opus 5 on release evals — alifcoder · 2026-11-06
- Klang AI open-sources Pianissimo-da, claiming SOTA Danish speech recognition — jaugjaug · 2026-11-06
- TailSteer: Drift-Resistant Multi-Turn Image Editing from CUHK MMLab — MMLab-CUHK · 2026-11-06
- Semir's goose down TVC generated with Opus 5.5 and Seedance 2.5 — ring_hyacinth · 2026-11-06
- Midjourney prompt builds a believable candid breakfast scene, full prompt shared — tisch_eins · 2026-11-06