1.2K views · 22 likes · 1 reposts (at collection)
Nathan Lambert compares DeepSeek's reported inference volume with four providers
- Who
- Nathan Lambert
- What
- reconstructed inference volume reports comparing DeepSeek's daily token volume with open model inference on other platforms
What happened
Nathan Lambert says DeepSeek serves 150T tokens a day on one cluster, more than open model inference on OpenRouter, Fireworks, Together, and BaseTen combined. He presents his reconstruction of inference volume reports across those companies and quotes @teortaxesTex, who links to DeepSeek's deployment documentation and notes that the number of clusters is unknown. https://github.com/deepseek-ai/DeepInfer/blob/main/doc/inference_deployment_qa.md https://twitter.com/teortaxesTex/status/2113685096389579101
Read next
See the full storyNathan Lambert Says DeepSeek’s Daily Token Volume Exceeds Four Providers CombinedThe story so far
Nathan Lambert reconstructed reported inference volumes and said DeepSeek serves 150T tokens per day on one cluster, exceeding the combined open model inference volume on OpenRouter, Fireworks, Together, and BaseTen. Lambert quoted @teortaxesTex, who linked to DeepSeek’s deployment documentation and noted that the number of clusters is unknown. The 150T-token figure therefore applies to the single cluster cited in the comparison, rather than an established total across DeepSeek’s clusters.
Keep up with AI, one email a day
Model releases, price changes and compute developments, with original sources. Free, confirmed by email, and easy to unsubscribe. Read a past issue · 2026-10-23