Ling 3.0 Tiny

At Q4_K_M, Ling 3.0 Tiny needs ~6.3GB of usable memory (weights + KV cache + runtime). In practice that means a 12GB+ Android phone or a 12GB+ iPhone — iOS lets a single app use less of its RAM than Android does.

On the Galaxy S25 Ultra (Snapdragon 8 Elite) it runs at ~39.8 tokens/s at Q4_K_M, estimated from memory bandwidth. Anything above ~8 tokens/s feels smooth for chat.

Yes — the iPhone Air runs it at ~39.8 tokens/s at Q4_K_M, using 6.3GB of its ~7.8GB usable memory.

Q4_K_M (4.9GB download) is the size/quality sweet spot of the 9 quants available. Total memory needed is ~6.3GB once the KV cache and runtime are counted — the download size alone understates it.

105 of the 139 phones we track have enough estimated usable memory at Q4_K_M; 105 also reach our usable-speed threshold of 3 tokens/s, and 105 run smoothly. Memory-fit alone does not mean a good experience.

54 of the 54 exact Mac configurations we track have enough estimated working memory. Each Mac uses the recommended Q4_K_M when it fits, or the largest smaller tracked quant that fits, and separates installed unified memory, the macOS reserve, and estimated speed.

It's tagged for chat, coding, reasoning. At 7.9B parameters (1.4B active — it's a MoE, so it decodes faster than its size suggests), it balances quality and speed for daily use.