← All stories
AI & Tech

Chamath's Testing Shows Open Source Chinese Models Are 16x Cheaper Than Anthropic

All-In Podcast · AI Sovereignty Wars, Palantir-Nvidia Deal, SCOTUS Birthright Ruling, Newsom’s CA Budget Lie · July 3, 2026
Chamath's Testing Shows Open Source Chinese Models Are 16x Cheaper Than Anthropic
All-In Podcast
All-In Podcast
AI Sovereignty Wars, Palantir-Nvidia Deal, SCOTUS Birthright Ruling, Newsom’s CA Budget Lie
"When you use our harness with Claude, it was simultaneously 1.4x cheaper and 1.5x faster than just using Anthropic Opus-48 alone. But if you wrap the open source model with our software factory, it was 16.4x cheaper. Now it was 3 times slower, but you know, you're talking about a couple of extra hours to save 16.4x."
Chamath revealed that 8090's Software Factory running on Chinese open source models delivers enterprise AI tasks at 16.4x lower cost than Anthropic, though 3x slower. He argued that at this price differential, continuing to use frontier labs that leak your competitive advantage to competitors is derelict and irresponsible for enterprises. The data supports the case for AI sovereignty using open models.

About this episode

The All-In Podcast hosts Jason Calacanis, David Sacks, Chamath Palihapitiya, and David Friedberg dedicated their episode to the escalating conflict over AI sovereignty and enterprise data control. The discussion was sparked by Palantir CEO Alex Karp's CNBC interview attacking frontier AI labs, where he declared enterprises have lost trust in companies like Anthropic and OpenAI over fears they are harvesting customer data and intellectual property to build competing products. Sacks revealed that Anthropic blindsided its partner Figma by launching Claude Design while an Anthropic executive sat on Figma's board, exemplifying a pattern where the company launches vertical apps in categories created by its own customers. Chamath presented data from 8090 showing Chinese open source models are 16.4 times cheaper than Anthropic for enterprise tasks, arguing that continuing to use frontier labs that leak competitive advantages is now irresponsible. The hosts reached consensus that enterprises must adopt AI sovereignty strategies using open source models and on-premise infrastructure to protect their intellectual property, with Friedberg predicting a fundamental shift from centralized cloud AI to distributed enterprise-owned systems. The episode also covered the Supreme Court's 6-3 decision upholding birthright citizenship under the 14th Amendment, striking down Trump's executive order in a major legal defeat. Friedberg delivered an extensive analysis of California's fiscal crisis, revealing $1.5 to $2 trillion in unaccounted pension and healthcare liabilities beyond the state's ballooning $355 billion budget, predicting defaults that could require federal bailouts and trigger a constitutional crisis. He controversially predicted AOC would become president and that red states would refuse to bail out California's mismanagement, potentially threatening the union. The hosts debated immigration policy with Friedberg advocating for a merit-based system favoring productive makers over those seeking benefits, while disagreeing on the timeline and extent of AI-driven job displacement.

Key takeaways

More stories More from All-In Podcast