Written by Jakub Rusinowski · Last updated September 8, 2026
Model library → Nex-N2 → Nex-N2 mini
35B total, 3B active, Apache 2.0 — 21.9 GB of weights at Q4_K_M and 23.8 GB with an 8K context, so it lands tight on a 24 GB card and clears a 32 GB one. Scores 74.4 on SWE-bench Verified and 60.7 on Terminal-Bench 2.1, which puts a genuinely agentic model inside consumer memory. Sparsity is the whole trick: VRAM is charged on all 35B, decode speed on just 3B.
Nex-N2 mini needs about 22 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.
| Parameters | 35 Billion (3B active) |
| Context window | 262,144 |
| Architecture | Mixture-of-Experts (Qwen3.5-35B-A3B-Base post-train) |
| Provider | Nex-AGI |
| Licence | Apache 2.0 |
| Specified at | Q4_K_M |
| System RAM | 32 GB |
| Record updated | 2026-09-08 |
Apache-2.0 — commercial use permitted. Commercial use permitted. No usage restrictions beyond attribution.
Modelled on a reference NVIDIA RTX 4090 (24 GB), with no KV cache (this record has no published architecture). Speed figures are ESTIMATES from the memory-bandwidth roofline described on the methodology page, not benchmarks we ran — rows marked measured come from published or reader-submitted runs. VRAM here includes the KV cache, so it reads higher than the headline figure above, which does not.
| Quant | Weights | VRAM needed | Est. speed | Fit on 24 GB |
|---|---|---|---|---|
| Q2_K | 11.5 GB | 12.3 GB | ~209 tok/s (est.) | Fits comfortably |
| Q3_K_M | 14.9 GB | 15.7 GB | ~193 tok/s (est.) | Fits comfortably |
| Q4_K_M | 21.1 GB | 21.9 GB | ~170 tok/s (est.) | Tight fit |
| Q5_K_M | 24.8 GB | 25.6 GB | ~24 tok/s (est.) | Offloads to system RAM (slow) |
| Q6_K | 28.7 GB | 29.5 GB | ~22 tok/s (est.) | Offloads to system RAM (slow) |
| Q8_0 | 37.2 GB | 38.0 GB | ~19 tok/s (est.) | Offloads to system RAM (slow) |
| F16 | 70.0 GB | 70.8 GB | — | Won't fit |
Want the memory numbers alone, at every quantization level and your own context length? Use the Nex-N2 mini VRAM calculator.
or compare on Vast.ai from $0.35/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
The cheapest catalogued GPU that runs Nex-N2 mini is the AMD Radeon RX 7900 XTX (24 GB).
Install Ollama, then run:
ollama run nex-n2Weights on Hugging Face: nex-agi/Nex-N2-mini.
Best for: agentic tasks, coding, consumer gpu, software engineering.
← All Nex-N2 models | VRAM calculator | Check your own hardware