AI inference pricing

Knowledge graph concept · also referred to as charge per tokens, charge for inference, AI compute, agent-hours, API prices, cost per token, serving economics, inference pricing, ChatGPT Pro, pricing scheme, unit economics, token usage, costing higher, cost a ton on the API, cheaper than a big token sub, API charges, input tokens, output tokens, cheap to run, pay premium, price of electricity, token pricing

Mentioned in

• The barbell effect will deprioritize mid-sized models X thread · 2026-09-18
• Where DeepSeek V4.1 Pro could land X thread · 2026-09-16
• DeepSeek V4.1 Flash plays Catan against GPT-5.6 X thread · 2026-09-15
• DeepSeek is winning on unit economics X thread · 2026-09-10
• The ideal model could run locally X thread · 2026-08-14
• Write-up of Reiner Pope's Lecture: How GPT, Claude, and Gemini Are Actually Trained and Served post · 2026-07-14
• The localening is here X thread · 2026-06-17
• Stop defaulting to weak coding agents for serious work X thread · 2026-01-17
• Anthropic pricing discourages agent-native development X thread · 2026-01-10
• Anthropic's pricing is stupid log · 2026-01-10
• GPT-5.2 xhigh feels like a careful systems debugger X thread · 2025-12-31
• Monetize AI, not the editor post · 2025-01-26

Related concepts

• AI model competition 8
• Open-weight language model 6
• AI coding agent 3
• Inference-time computation 3
• Model size 3
• Agent harness 2
• AI benchmark 2
• AI model reliability 2