Quote · Moonshots with Peter Diamandis
Google’s Jeff Dean Exits, SpaceX Hits $100B in Rev & OpenAI’s Astra Solves Decade-Old Math Problems with Emad Mostaque | EP #277
Where this was said
Alibaba's Qwen 3.8 Max: Open-Weight Frontier at 88% Less Cost
At 1:04:00 · chapter starts 1:00:25
Peter Diamandis walks through Alibaba's Qwen 3.8 Max release: a multimodal model with 2.4 trillion total parameters, 95 billion active per request, and a million-token context window, capable of processing 100-hour videos and building apps from screenshots [1] — Peter Diamandis "Qwen 3.8 Max: 88% cheaper than Claude: Alibaba's Qwen 3.8 Max model is priced 88% below Claude Fable 5 and 80% below GPT-5.6 SOL, while ran…" 1:00:46 . At $2 per million input tokens and $6 per million output tokens, it's 88% cheaper than Claude Fable 5 — and Alibaba's stock responded with a 7% gain. Alex Wissner-Gross delivers his now-famous framing: the Chinese Communist Party is ironically saving American capitalism from itself. Without Qwen and Kimi applying competitive pressure, Western frontier labs would have every incentive to restrict access and optimize margins rather than compete on capabilities. Emad Mostaque adds a key insight: the more strategically important release is Qwen 27B — a model that fits on a 16GB RAM MacBook and is approaching cyber-attack capability thresholds, raising national security concerns. Salim Ismail echoes the open-source forcing function thesis but notes these models still require substantial infrastructure to self-host.
Alibaba's Qwen 3.8 Max model is priced 88% below Claude Fable 5 and 80% below GPT-5.6 SOL, while ranking roughly third or fourth globally on capability benchmarks.
Without competitive pressure from Alibaba's Qwen and Moonshot's Kimi series, Western frontier labs would have no incentive to compete on cost or capability. Alex Wissner-Gross argues that Chinese open-weight models are a space-race level forcing function that is paradoxically the best possible scenario for American AI competitiveness.
Peter Diamandis estimated that a new frontier AI model is being released on average every 5.5 days, reflecting the accelerating pace of AI development.
