OpenRouter CEO: Why Chinese Open Models Are Beating the US | Why Enterprises Fear OpenAI & Anthropic
- Source
- youtube.com
- Author
- 20VC
- Date
Context
Alex Atallah, co-founder and CEO of OpenRouter, a gateway that lets developers call many AI models through one interface, told 20VC that an unexpected layer of companies emerged to models, meaning models whose trained parameters are published. Early on, it was unclear whether the three big cloud providers would end up serving all of them. Atallah argues people rarely run GLM on a hyperscaler and instead use providers such as Fireworks and Together AI, which Atallah credits with hosting new models faster and handling edge cases better.
That matters to OpenRouter because, in Atallah's account, the same can perform differently depending on the host. OpenRouter says it benchmarks models across providers continuously, sees differing results that change over time, and sends more traffic to a provider soon after detecting a quality gain, a speedup or a price cut. This is Atallah's description of OpenRouter's own product, not an independent measurement.
- Host — The computer on which a process runs.
- Open weights — A model whose learned parameter values have been made available to download, under terms the publisher sets.
- Inference — Using a trained model to produce an output from a supplied input.
- Model — A computational system that uses learned patterns to produce an output from information supplied to it.
videoReplit CEO: Why the SaaS Apocalypse is Justified & Why Coding Models are Plateauing | Amjad Masad
videoCerebras CEO on the Future of Data Centres, Token Costs & Memory | Should US Companies Sell to China
videoInside Legora's Tech Stack: Why Token Maxing is Failing Enterprise Startups | Legora CTO
videoAndrew Feldman, Cerebras Co-Founder and CEO: The AI Chip Wars & The Plan to Break Nvidia's Dominance
Checking sign-in…
Loading comments…


