OpenRouter

More Trillion Dollar IPOs, Anthropic $3T, Zuck’s Price War, China Ends Open Source?, Trump Accounts adds a buyer-side cost-cutting anecdote. Chamath Palihapitiya says his team used OpenRouter, GLM 5.2, and other routing choices to cut model costs by roughly 95%, turning OpenRouter from a general model-marketplace example into a concrete Enterprise AI ROI Audit and Model Routing Cost Control tool.

OpenRouter is the model-routing and API aggregation company discussed in E246｜何谓蒸馏？聊聊硅谷如何看中国开放模型逼近前沿. Keith Zhai uses it as an example of an intermediary that benefits when strong open and closed models coexist, because customers have more reason to compare and route across models rather than defaulting to one frontier provider.

For the wiki, OpenRouter belongs in Model Routing Cost Control and AI Inference Cost Structure. Its value increases when open models such as Kimi K3 make model choice a routine product decision across price, latency, quality, policy, and deployment constraints.

Featherless AI: When Your Weekend Experiment Makes More Than Your Startup adds a boundary from the provider side. Eugene Chia says Featherless AI can look OpenRouter-like to users because both promise easy access to many models, but Featherless hosts models directly while OpenRouter routes requests to providers such as Featherless.

Vol. 172 Codex 卖重置套餐，DeepSeek 峰谷调价，苹果重回 5 万亿等 adds a heavy-user market-signal version. The hosts describe OpenRouter as the “middle station” that lowers switching cost across a growing model set and as a place to observe which models users actually pay for when DeepSeek, OpenAI, Anthropic, Kimi, and other providers move prices, limits, or capabilities.

Connections

• Model Routing Cost Control - user/product-level routing practice OpenRouter exemplifies.
• AI Inference Cost Structure - token-price and provider-cost layer that makes routing valuable.
• Open Source AI Models, Kimi K3, and Closed Model API Moat Pressure - model diversity and API moat pressure behind its opportunity.
• Agent Inference Workload - agent workloads can make routing more valuable because long prompts, cache reuse, and repeated steps create cost differences.
• Featherless AI, Long-Tail Model Hosting, and GPU Hot Swapping - hosted-provider layer clarified by the Featherless episode.
• DeepSeek, OpenAI, Anthropic, Kimi, Peak-Valley AI Inference Pricing, and AI Subscription Economics — Vol. 172’s model-choice and routing-market branch.