Tech Tactician
地端 AI 實驗室

Podcast

地端 AI 實驗室

每天精選 1~3 個可以下載到自己機器上跑的開源模型與 LoRA，用輕鬆對話帶你掌握它能用在哪些場景、你的顯卡跑不跑得動、跟同類差在哪。本節目為情報策展，功能與硬體規格一律引述官方 model card，未經我方實測，更多內容請見 (https://blog.markkulab.net)

17 episodes

Apple PodcastRSS Feed
🧪 Challenging 512K Context! Nanbeige4.3-33B Becomes the Strongest Local Agent: Local AI Lab

🧪 Challenging 512K Context! Nanbeige4.3-33B Becomes the Strongest Local Agent: Local AI Lab

Mark Ku·

🧪 1.3B Runs on Laptops: MiniCPM-OCR-1.0 Turns Local Invoices into JSON in Seconds: Local AI Lab

🧪 1.3B Runs on Laptops: MiniCPM-OCR-1.0 Turns Local Invoices into JSON in Seconds: Local AI Lab

Mark Ku·

🧪 64MB Ultra-Lightweight phonon-31m-en Beats Kokoro in Blind Test: Local AI Lab

🧪 64MB Ultra-Lightweight phonon-31m-en Beats Kokoro in Blind Test: Local AI Lab

Mark Ku·

🧪 M4 Pro Blazes Past 125 Tokens! Offline Coding Agent Model flint Debuts: Local AI Lab

🧪 M4 Pro Blazes Past 125 Tokens! Offline Coding Agent Model flint Debuts: Local AI Lab

Mark Ku·

🧪 66x KV Cache Savings! MemT-1.0-Nano Enables Local Native Long-Term Memory | Local AI Lab

🧪 66x KV Cache Savings! MemT-1.0-Nano Enables Local Native Long-Term Memory | Local AI Lab

Mark Ku·

🧪 Can 4B Parameters Control Thinking Budget? Pair with an 8-Step Distilled Image Gen Companion to Run Agents Offline: Local AI Lab

🧪 Can 4B Parameters Control Thinking Budget? Pair with an 8-Step Distilled Image Gen Companion to Run Agents Offline: Local AI Lab

Mark Ku·

🧪 GPU-Free and 15x Faster: 1.25M Parameter indra-ocr is a Local Savior | Local AI Lab

🧪 GPU-Free and 15x Faster: 1.25M Parameter indra-ocr is a Local Savior | Local AI Lab

Mark Ku·

🧪 322M beats 28B? convaiinnovations/laya claims to be 7.8x faster than the competition｜Local AI Lab

🧪 322M beats 28B? convaiinnovations/laya claims to be 7.8x faster than the competition｜Local AI Lab

Mark Ku·

🧪 The most eye-popping download numbers this episode aren't a chat model, they're a decision LoRA: how GEV-26B-Decide lets the agent decide for itself whether to think more｜Local AI Lab

🧪 The most eye-popping download numbers this episode aren't a chat model, they're a decision LoRA: how GEV-26B-Decide lets the agent decide for itself whether to think more｜Local AI Lab

Mark Ku·

🧪 The 740M embeddinggemma-2 handles text, images, video, and audio, and it even fits on your phone | Local AI Lab

🧪 The 740M embeddinggemma-2 handles text, images, video, and audio, and it even fits on your phone | Local AI Lab

Mark Ku·

177B Model on a Diet: GSQ-RCO Quantized Version Matches Original with 2.2M Downloads

177B Model on a Diet: GSQ-RCO Quantized Version Matches Original with 2.2M Downloads

Mark Ku·

🧪 164 MB, 174x Real-Time: Phonon-2 Makes Speech-to-Text Free From Cloud Taxes | Local AI Lab

🧪 164 MB, 174x Real-Time: Phonon-2 Makes Speech-to-Text Free From Cloud Taxes | Local AI Lab

Mark Ku·

探索十大熱門分類

#1
AI

AI

35 articles
#2
DevOps

DevOps

32 articles
#3
Frontend

Frontend

39 articles
Management

Management

25 articles
Infra

Infra

22 articles
PowerShell

PowerShell

15 articles
Travel

Travel

12 articles
E-Commerce

E-Commerce

11 articles
.NET Core

.NET Core

11 articles
Cloud

Cloud

10 articles

Subscribe to Newsletter

Subscribe to get new posts delivered instantly — never miss a tech share.

By submitting, you agree to receive emails. You can anytime.