Foundation AI

Antares 350M

Name: Antares 350M
Rating: 50 (196 reviews)
Author: Foundation AI

当前

HuggingFace

1.2K下载量50点赞Jul 2026发布日期33K tokens上下文Apache 2.0许可证31 基础质量

Antares 350M (0.3499999940395355B parameters) requires approximately 2.4 GB of VRAM with Q4_K_M quantization. For the best balance of quality and speed, we recommend hardware with at least 3 GB of VRAM.

快速开始

— 复制粘贴即可本地运行

Copy-paste commands to run Antares 350M on your machine.

Run

docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
  --hf-repo "fdtn-ai/antares-350m" \
  --hf-file "antares-350m-Q4_K_M.gguf" \
  -c 4096 -ngl 99

Quick specs

Parameters0.35B

Architecturedense

Context33K tokens

Modalitytext

Min RAM0.1 GB

Rec. RAM0.2 GB (Q4_K_M)

LicenseApache 2.0

FamilyAntares

About this model

Antares 350M is the smallest member of Foundation AI's security-focused Antares family, post-trained on IBM's Granite 4.0 350M. At 0.35B parameters it is designed for embedded and always-on security agent workloads that must run on CPU with a minimal memory budget.

•0.35B parameters — under 300MB at Q4, runs on almost anything.
•Security and vulnerability-detection focus, same recipe as Antares 1B.
•Built on IBM Granite 4.0 350M.
•32,768-token context.
•Apache 2.0 licensed.

Antares 350M 的最佳选择

GTX 1650 4GBB

4 GB

RTX 3050 Ti Laptop 4GBB

运行此模型

Antares 350M on GTX 1650 4GB Antares 350M on RTX 3050 Ti Laptop 4GB Antares 350M on Intel Arc A370M 4GB

量化选项

各量化级别的 VRAM 估算

No hardware detected — fit column shows raw VRAM estimates

Quant	Bits	VRAM	Quality	Fit
Q1_0_G128	1.125	0.1 GB	Very Low	—
Q2_0_G128	1.71	0.1 GB	Low	—
Q2_K	2	0.1 GB	Low	—
Q3_K_S	3	0.2 GB	Low	—
NVFP4	4	0.2 GB	Medium	—
Q4_K_M	4	0.2 GB	Medium	—
Q5_K_M	5	0.3 GB	High	—
Q6_K	6	0.3 GB	High	—
Q8_0	8	0.4 GB	Very High	—
F16	16	0.7 GB	Maximum	—

硬件兼容性

全部硬件的适配估算

打开计算器

Computing compatibility...

内存详细分析

Reference: RTX 2060 6GB

Weights0.2 GB

KV Cache0.4 GB

Runtime1.2 GB

Headroom0.6 GB

常见问题

FAQ — Antares 350M

另请参阅

量化指南评分方法打开计算器