Cloud vs. Rig TCO
Plot cumulative 36-month API spending against hourly RunPod/H100 rentals and owned RTX/Mac hardware with electricity and resale value.
Size daily PDF ingestion pipelines across 1M-token contexts with implicit/explicit caching.
Multi-turn history expands average input context to 8,000 tokens/req. Output accounts for 25% of total spend.
| Model | Delta | |||
|---|---|---|---|---|
| Claude Haiku 5.5 (High-Volume)Anthropic | $0.10 | $0.50 | $522.76 | Save $3,398 |
| GPT-4.1 nano (Ultra-Low Cost)OpenAI | $0.10 | $0.40 | $532.66 | Save $3,388 |
| DeepSeek V4.1 Flash (Off-Peak)DeepSeek | $0.15 | $0.60 | $714.98 | Save $3,206 |
| GPT-4o mini (Legacy Budget)OpenAI | $0.15 | $0.60 | $890.3 | Save $3,030 |
| Gemini 3.5 Flash-Lite (Budget Edge)Google | $0.30 | $2.50 | $1,834.62 | Save $2,086 |
| GPT-4.1 mini (Balanced Fast)OpenAI | $0.40 | $1.60 | $2,130.63 | Save $1,790 |
| DeepSeek V4 Pro (Off-Peak Reasoning)DeepSeek | $0.66 | $1.98 | $2,991.55 | Save $929 |
| Llama 3.3 / 4 70B-Class (Hosted API)Open-Weight Host | $0.59 | $0.79 | $3,802.02 | Save $119 |
| Gemini 3.8 Flash (Intro Paid Tier)Google | $0.75 | $3.75 | $3,920.73 | Active Baseline |
| o3-mini (Reasoning)OpenAI | $1.10 | $4.40 | $6,528.84 | +$2,608 |
| GPT-5 (Flagship Reasoning)OpenAI | $1.25 | $10.00 | $7,533.28 | +$3,613 |
| Gemini 4 HydrogenGoogle | $1.50 | $7.50 | $7,658.84 | +$3,738 |
| Claude Sonnet 5.5 (Coding & Agents)Anthropic | $2.00 | $10.00 | $10,211.78 | +$6,291 |
| GPT-6.2 SolOpenAI | $2.00 | $10.00 | $10,211.78 | +$6,291 |
| Gemini 4 CarbonGoogle | $2.00 | $10.00 | $10,211.78 | +$6,291 |
| GPT-4.1 (1M Long-Context)OpenAI | $2.00 | $8.00 | $10,653.13 | +$6,732 |
| Gemini 3.1 Pro (Long-Context)Google | $2.00 | $12.00 | $10,987.94 | +$7,067 |
| GPT-4o (Legacy Omni)OpenAI | $2.50 | $10.00 | $14,838.28 | +$10,918 |
| Gemini 4 NeonGoogle | $3.00 | $15.00 | $15,317.67 | +$11,397 |
| Claude Opus 5.5 (Agentic Frontier)Anthropic | $4.00 | $20.00 | $20,423.56 | +$16,503 |
| Claude Fable 5.2Anthropic | $8.00 | $40.00 | $40,360.13 | +$36,439 |
| GPT-6.2 AstraOpenAI | $8.00 | $40.00 | $40,847.13 | +$36,926 |
Single-turn structured extraction keeps output tokens tight and margins predictable. Pre-populated at 25,000 daily requests with 8,000 input tokens, 350 output tokens, 40% prompt caching, and 1 turns/session.
Primary Technical & US Grid Benchmarks
Plot cumulative 36-month API spending against hourly RunPod/H100 rentals and owned RTX/Mac hardware with electricity and resale value.
Count live prompt tokens across English prose and JSON/code, size RAG document chunks, and compute float32 vs. int8 HNSW vector RAM.
Test Llama 70B, Qwen 32B, or DeepSeek R1 across RTX 3090/4090/5090 & Apple Silicon Macs with GQA KV-cache math and tokens/sec speed.