Shanghai cafes now serve free local DeepSeek v4 flash and MiniMax H3 via in-store inference

赛博禅心 · wechat · 2026-09-13

After Beijing, two Shanghai cafes now offer free DeepSeek v4 flash and MiniMax H3 access. Technical details:

• Models run locally in Beijing on shared NVIDIA DGX hardware (100+ tokens/sec)
• The two stores are linked via an intranet bridging their WiFi networks, plus a token-distribution relay system
• Usage: connect to store WiFi, set Base to https://token.agi.bar/v1 with API key sk-agibar-static
• Unlimited free MiniMax H3 included, with a curl example for video generation (width, fps, duration parameters)

Original post →

More from Infra

• New research: standard SGD matches AdamW for LLM RL training, with far less memory overhead — zhaoran_wang · 2026-09-14
• RTX 5090 stock nearly gone as prices spiral further — DustNearby2848 · 2026-09-14
• OneLA Shares Linear-Attention States Across Beams for 1.54-2.46x Faster Generative-Rec Decoding — _reachsumit · 2026-09-14
• UBTech Opens World's First 10,000-Unit Humanoid Robot Factory; SK's AI Data Center Hits 900MW — 创业邦 · 2026-09-14
• Deep dive: thermal gradients are HBM's #1 scaling bottleneck, says new architecture breakdown — blaizedsouza · 2026-09-14
• Perplexity launches Hybrid Compute to split AI tasks between cloud and local Mac — Aiden_Tech_Ai · 2026-09-14

Infra channel →