Power user burns 26.2B tokens in a month for $94 on DeepSeek agents
teortaxesTex · x · 2026-11-22
Account teortaxesTex shared his "Context Insights" dashboard for running agents on DeepSeek via opencode-go, revealing the real cost structure of extreme usage:
- 379 sessions, 26.2B tokens, estimated cost just $94.45, with 569h active time and 98.4k model calls;
- Cache hit rate of 98.53% is the key to keeping costs low;
- Token mix: Tool Calls 42.9%, Reasoning 31.9%, Tool Schemas 19.1%, while actual answers are only 1.4% — most agent tokens go to tool orchestration and reasoning, not output.
He extrapolates that market-wide adoption at this intensity would need 116M GPUs and 240 GW of power, but argues cheaper, denser models (like Muse Ember at $0.15/test, the fastest on the Vals Index, and tiny tool-augmented models) keep invalidating "we'll run out of compute" conclusions.
More from Infra
- AI skeptic: market crash could permanently erase AI stock premium, then a supply glut hits — teortaxesTex · 2026-11-22
- Gary Marcus Amplifies Claim That 'Bank of Nvidia' Is Propping Up Token Prices — GaryMarcus · 2026-11-22
- AWS Marketplace adds x402: agents pay per request straight to your AWS bill — MurrLincoln · 2026-11-22
- Burkov: open-weight models will soon run as cheap as a cloud VPS as AI demand cools — burkov · 2026-11-22
- Samsung's memory capacity for 2027-2029 is essentially sold out; customers seek supply deals through 2031 — sudoraohacker · 2026-11-22
- TensorFold 1.3.3 ships 124 PRs in 24h, boosts GLM decode to 172 tok/s on GB10s — AIFlow_ML · 2026-11-22