Nex-AGI35B (3B active)Q4_K_M 下约 22 GB 显存

Nex-N2.5 mini — 显存、速度与本地部署

作者： Jakub Rusinowski · 最后更新：

The only Nex model a consumer card holds at all: 21.9 GB of weights at Q4_K_M, 23.8 GB once an 8K KV cache is added — tight on a 24 GB 4090, comfortable on a 32 GB 5090. Reads 3B parameters per token, which is why a 35B model decodes at small-model speed. Scores 73.4 on Terminal-Bench 2.1 and 83.4 on BrowseComp — the latter within 10 points of the 1.6T Max. Parameter counts are inherited from the Nex-N2 base rather than published for N2.5 directly.

Nex-N2.5 mini 在 Q4_K_M 下约需 22 GB 显存——量化权重加框架开销，不含 KV 缓存。在 Apple Silicon 上，这部分来自统一内存。

逻辑84
创意78
编程85

按量化级别的显存与速度

计算基准：NVIDIA RTX 4090 (24 GB)。仅含权重与开销：该模型架构未公开，因此未计入 KV 缓存。

量化显存速度（估算）适配
Q2_K
2.63 bpw
12.3 GB~209 tok/s可运行
Q3_K_M
3.41 bpw
15.7 GB~193 tok/s可运行
Q4_K_M
4.83 bpw
21.9 GB~170 tok/s勉强
Q5_K_M
5.67 bpw
25.6 GB~24 tok/s需卸载
Q6_K
6.56 bpw
29.5 GB~22 tok/s需卸载
Q8_0
8.50 bpw
38 GB~19 tok/s需卸载
F16
16.00 bpw
70.8 GB—放不下

黑色标记 = NVIDIA RTX 4090 (24 GB) 上的可用显存。 估算来自内存带宽屋顶线模型，详见 方法说明页. Nex-N2.5 mini 显存计算器 →

运行 Nex-N2.5 mini

目录中能运行 Nex-N2.5 mini 的最便宜 GPU 是 NVIDIA GeForce RTX 5090 (32 GB).

购买此硬件 NVIDIA GeForce RTX 5090 32GB — 32 GB VRAM · 575 W board power立即云端部署 RunPod 上的 RTX 4090 — 低至 $0.34/小时 · 价格核实于 2026-07

或在 Vast.ai 比较，低至 $0.35/小时 (typical low · varies)

作为亚马逊联盟成员，我们从符合条件的购买中获得收入。云 GPU 链接为推荐链接——我们可能获得佣金，您无需额外付费。

联盟营销声明: 本页部分链接为联盟推广链接——如果你通过它们购买，LLM Configurator 可能会获得佣金，而你无需支付任何额外费用。作为亚马逊联盟成员（Amazon Associate），LLM Configurator 会从符合条件的购买中获得收益。
NVIDIA GeForce RTX 5090 32GB
32 GB VRAM · 575 W board power
2026年价格波动较大——请以当前商品页价格为准。

如何运行 Nex-N2.5 mini

安装 Ollama，然后运行：

ollama run nex-n2-5
Hugging Face 上的权重： nex-agi/Nex-N2.5-mini ↗

规格

Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.

Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.

参数量
35 Billion (3B active)
上下文窗口
262,144
架构
Mixture-of-Experts (Qwen3.5-35B-A3B base, multimodal)
提供商
Nex-AGI
许可证
Open-weight (terms unpublished)
Practical deployment
Single consumer GPU
规格量化
Q4_K_M
系统内存
32 GB
记录更新于
2026-09-08
许可证Custom Open-Weight允许商业使用

Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.

质量与使用场景

评分由模型作者或独立评测方发布——衡量质量而非吞吐量，并非我们实测。

最适合agentic taskscodingconsumer gpumultimodal

我的 GPU 能运行 Nex-N2.5 mini 吗？

Nex-N2.5 的其他尺寸

Nex-N2.5 mini — 常见问题

How much VRAM does Nex-N2.5 mini need?

About 22 GB at Q4_K_M — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.

Does Nex-N2.5 mini run on an RTX 4090 (24 GB)?

Yes. Nex-N2.5 mini needs about 22 GB at Q4_K_M, inside a 24 GB card, at an estimated 170 tokens/sec.

How do I run Nex-N2.5 mini locally?

Install Ollama and run `ollama run nex-n2-5`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.

What other sizes does Nex-N2.5 come in?

Nex-N2.5 mini (22 GB), Nex-N2.5 Pro (240 GB), Nex-N2.5 Max (967 GB). Every size shares the family's training and licence; the larger ones score higher and need proportionally more memory.