惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

小众软件
小众软件
博客园_首页
博客园 - 聂微东
T
Tailwind CSS Blog
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
J
Java Code Geeks
The Cloudflare Blog
aimingoo的专栏
aimingoo的专栏
Martin Fowler
Martin Fowler
D
Docker
人人都是产品经理
人人都是产品经理
WordPress大学
WordPress大学
博客园 - 三生石上(FineUI控件)
Microsoft Azure Blog
Microsoft Azure Blog
Recent Announcements
Recent Announcements
Apple Machine Learning Research
Apple Machine Learning Research
阮一峰的网络日志
阮一峰的网络日志
B
Blog RSS Feed
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
Microsoft Security Blog
Microsoft Security Blog
L
LangChain Blog
Jina AI
Jina AI
博客园 - Franky
D
DataBreaches.Net

Hacker News - Newest: "AI"

AI can't read an investor deck AI as an attorney? Student uses ChatGPT, Gemini to sue UW over alleged racial discrimination Hacking MCP Servers in AI Systems – The Rug Pull: Tool Changes After Approval GitHub - MeepCastana/KubeezCut: Free Web based video editor Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers Coming soon: 10 Things That Matter in AI Right Now DARPA built an AI to fact-check enemy weapons claims What explains heterogeneity in AI adoption? When AI Meets Muscle: Context-Aware Electrical Stimulation Promises a New Way to Guide Human Movements - Department of Computer Science AI Changed How We Build. It Did Not Change What Matters. Linux rules on using AI-generated code - Copilot is OK, but humans must take 'full responsibility for the… Meta spins up AI version of Mark Zuckerberg to engage with employees Code Mode: Let Your AI Write Programs, Not Just Call Tools | TanStack Blog GitHub - Delavalom/graft: Go framework for building AI agents. Type-safe tools, multi-provider (OpenAI, Anthropic, Gemini, Bedrock), zero vendor SDKs. India's TCS tops estimates, says new AI models did not dent services demand Gen Z's fading AI hype Strong feeling: we are in a folded AI reality GitHub - machinarii/total-recall-catalog: A reference catalog of latest knowledge retrieval, memory & RAG systems GitHub - mensfeld/code-on-incus: Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically.. Quantization, LoRA, and the 8% Problem: Benchmarking Local LLMs for Production AI Iran war: We spoke to the man making Lego-style AI videos that experts say are powerful propaganda Powell, Bessent discussed Anthropic's Mythos AI cyber threat with major U.S. banks GitHub - immartian/bellamem: Persistent belief-graph memory for AI agents. Retrieves decisive context by importance — not recency, not RAG, not /compact. recursive-mode: The Repo-Native Operating System for AI Engineering After the attack on Sam Altman's home, will AI CEO's go on the offensive? The biggest advance in AI since the LLM Opus 4.6 vs GPT 5.4 One Prompt Unity World Generation Test “AI polls” are fake polls Client Challenge Can AI be a 'child of God'? Inside Anthropic's meeting with Christian leaders
Firefly AIBOX-K3 – An Edge AI mini PC powered by SpacemiT...
Debashis Das · 2026-05-12 · via Hacker News - Newest: "AI"

Back in July last year, SpacemiT unveiled the SpacemiT K3 SoC. After that, we saw some system information and early benchmarks come out around January this year. The company has just officially launched the K3 Pico-ITX SBC, which is now available through various distributors. Firefly has launched its own K3 hardware with the AIBOX-K3, a complete industrial-grade RISC-V edge computing box.

The AIBOX-K3 Edge AI mini PC is built around the SpacemIT Key Stone K3 octa-core processor and features an integrated AI engine that delivers up to 60 TOPS of compute performance, making it suitable for local LLM inference and edge AI applications.

AIBOX-K3 is a SpacemiT K3 Edge AI mini PC Box

Firefly AIBOX-K3 specifications:

  • SoC – SpacemiT K3
    • CPU
      • 8x 64-bit RISC-V X100 “big” cores clocked up to 2.4 GHz, RVA23 compliance; 130 KDMIPS performance (similar to RK3588)
      • 8x RISC-V A100 AI Cores with support for up to 1024-bit RVV1.0 parallel computing, optimized for matrix operations.
    • GPU – Imagination Technologies BXM4-64-MC1 GPU with Vulkan 1.3, OpenCL 3.0, and OpenGL ES 1.1/2.0/3.2 support
    • VPU
      • Video decoder – H.265, H.264, VP9 up to 4K @ 120 FPS
      • Video encoder – H.265, H.264 up to 4K @ 60 FPS
    • AI – Up to 60 TOPS (INT4) of AI performance using dedicated TCM and DMA acceleration channels
  • System Memory – 8GB, 16GB, or 32GB LPDDR5, 6400 MT/s (51GB/s bandwidth)
  • Storage
    • 128GB, 256GB, or 512GB UFS 2.2 on-board storage
    • M.2 socket for NVMe SSDs
  • Display – HDMI 2.0 port supporting up to 4K@60Hz
  • Networking – Dual Gigabit Ethernet (1000 Mbps) via RJ45 ports
  • USB
    • 2x USB 3.0 Type-A ports (Max: 1A)
    • 1x USB 3.0 DRD Type-C port (includes USB 2.0 OTG functionality)
    • 1x USB Type-C Console port for serial debug
  • Expansion – M.2 socket (PCIe 3.0 x4) supporting 2242 or 2280 NVMe SSDs or other compatible modules
  • Misc – Power button, Boot button, and Status LED
  • Power Supply – 9V to 20V  via a 5.5 x 2.1mm barrel jack (12V/5A recommended)
  • Dimensions – 93.4 x 93.4 x 50.0 mm
  • Weight – Approximately 500 grams
  • Temperature Range – Operating: -20°C to 60°C; storage: -20°C to 70°C
  • Humidity – 10%~90%RH (Non condensating)
  • Enclosure – Industrial-grade aluminum body with porous hexagonal structure and strip grille ventilation
AIBOX K3 Front Interface
AIBOX-K3 Front Interface
AIBOX K3 Back Interface
AIBOX-K3 Back Interface

The device delivers up to 60 TOPS (Sparse) of AI power and supports multimodal acceleration, enabling it to run large models with up to 30B parameters, including Mixture of Experts (MoE) models. According to Firefly, the system can achieve inference speeds of over 10 tokens per second when running a 30B model locally.

Security and virtualization are handled at the hardware level. The SoC uses a three-level M/S/U privilege architecture and includes protection against Specter and Meltdown–type attacks. It also supports PMP/ePMP and IOPMP for memory and I/O protection, as well as built-in encryption engines such as AES, SHA, RSA, SM2, SM3, and SM4.

In the download section, the company only provides basic Linux and Windows flashing tools. This means the operating system is not guaranteed to ship pre-installed, so the software may need to be installed separately. After looking further, I found that SpacemiT supports multiple operating systems on the K3 platform, including Bianbu OS 3.0, Ubuntu 26.04 (developed in collaboration with Canonical), OpenHarmony, OpenKylin, Fedora, Deepin, and others. The SpacemiT K3 support has already been added to Linux 7.0 (initial support only in mainline), and features such as Docker, KVM virtualization, and ROS are available. More information is available on SpacemIT GitHub account.

The Firefly SpacemiT K3-based AIBOX-K3 Edge AI mini PC starts at $349 on the company’ store in the 8GB RAM + 128GB storage configuration, while the 32GB RAM version costs $689.

AIBOX-K3 SpacemiT K3 Edge AI mini PC Dimension
Dimension

Debashis Das

Debashis Das is a technical content writer and embedded engineer with over five years of experience in the industry. With expertise in Embedded C, PCB Design, and SEO optimization, he effectively blends difficult technical topics with clear communication

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress. We also use affiliate links in articles to earn commissions if you make a purchase after clicking on those links.