Nativ(@Nativ_AI):Congrats to @Zai_org on GLM-5.3-Flash! Day 0 support in Nativ 🎉 320B params, 18B active · 1M native context · fully on-device. On an M3 Ultra (512GB) with Nativ v0.3.4, standard 4-bit MLX: ⚡ Up to 505 tok/s prefill · 32 tok/s decode 🚀 Batch 4 lifts decode to 70.9 tok/s 💾 Peak under 380GB, fully resident, no offload 📏 Benchmarked to 128K context A 320B MoE running entirely on your Mac. Try Nativ 👇

2026.08.26 18:08

Congrats to @Zai_org on GLM-5.3-Flash! Day 0 support in Nativ 🎉 320B params, 18B active · 1M native context · fully on-device. On an M3 Ultra (512GB) with Nativ v0.3.4, standard 4-bit MLX: ⚡ Up to 505 tok/s prefill · 32 tok/s decode 🚀 Batch 4 lifts decode to 70.9 tok/s 💾 Peak under 380GB, fully resident, no offload 📏 Benchmarked to 128K context A 320B MoE running entirely on your Mac. Try Nativ 👇