Can 2x RTX 3090 run Inkling 2 Mini?
Last updated:
Every measured result for Inkling 2 Mini on 2x RTX 3090
| Build | Quant / setup | Engine | tok/s | |
|---|---|---|---|---|
| 2x RTX 3090: Inkling 2 Mini W4A16 INT4 on vLLM TP2 + Ampere overlay, 136.6 tok/s bench (Tech2wild) | W4A16 INT4 + DSpark | vLLM TP2 + Ampere overlay, NVLink, 1 stream | 136.6 | source |
| 2x RTX 3090: Inkling 2 Mini W4A16 INT4 on vLLM TP2 + Ampere overlay, 136.6 tok/s bench (Tech2wild) | W4A16 INT4 + DSpark | vLLM TP2 + Ampere overlay, NVLink, 1 stream | 119 | source |
| 2x RTX 3090: Inkling 2 Mini W4A16 INT4 on vLLM TP2 + Ampere overlay, 136.6 tok/s bench (Tech2wild) | W4A16 INT4 + DSpark | vLLM TP2 + Ampere overlay, NVLink P2P off (NCCL_P2P_DISABLE=1), 1 stream | 100.5 | source |
| 2x RTX 3090: Inkling 2 Mini W4A16 INT4 on vLLM TP2 + Ampere overlay, 136.6 tok/s bench (Tech2wild) | W4A16 INT4 | vLLM TP2 + Ampere overlay, 1 stream | 89.8 | source |
Memory: 2x RTX 3090 has 48 GB. Inkling 2 Mini weights are about 20 GB in its native format (Thinking Machines Lab model card: 26.4B total / 3.4B active looped MoE (middle layers run 3 times, ~5.4B effective active), text + image in, up to 1M context, Apache 2.0, released 20 Nov 2026; one ~20 GB NVFP4/MXFP8 checkpoint). Arithmetic, not a measurement; long contexts need more. Check other quants in the VRAM calculator.
Speeds are decode (generation) tokens/sec as reported by each source; quant, engine and context differ, so small gaps are noise.
All measured speeds for Inkling 2 Mini: best home rig for Inkling 2 Mini.
Inkling 2 Mini on other hardware: Apple M3 Ultra · Apple M5 Ultra · DGX Spark (GB10) · RTX 3090 · RTX 4090
Other models measured on 2x RTX 3090: MiniCPM-V 4.7 · Nanbeige4.3 33B · Nex N2.5 mini · Qwen3.6 35B-A3B · Qwen3.8 27B
Referral link: if you sign up with Google and add $10, Runpod gives you and RigBase bonus credit (terms). It never affects our data.