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.
Account for silent chain-of-thought reasoning tokens (4x normal output length) across DeepSeek off-peak/peak API queues.
Multi-turn history expands average input context to 4,638 tokens/req. Output accounts for 72% of total spend.
| Model | Delta | |||
|---|---|---|---|---|
| GPT-4.1 nano (Ultra-Low Cost)OpenAI | $0.10 | $0.40 | $135.09 | Save $570 |
| Claude Haiku 5.5 (High-Volume)Anthropic | $0.10 | $0.50 | $158.75 | Save $546 |
| DeepSeek V4.1 Flash (Off-Peak)DeepSeek | $0.15 | $0.60 | $198.25 | Save $506 |
| GPT-4o mini (Legacy Budget)OpenAI | $0.15 | $0.60 | $207.4 | Save $497 |
| Llama 3.3 / 4 70B-Class (Hosted API)Open-Weight Host | $0.59 | $0.79 | $451.85 | Save $253 |
| GPT-4.1 mini (Balanced Fast)OpenAI | $0.40 | $1.60 | $540.37 | Save $164 |
| DeepSeek V4 Pro (Off-Peak Reasoning)DeepSeek | $0.66 | $1.98 | $704.69 | Active Baseline |
| Gemini 3.5 Flash-Lite (Budget Edge)Google | $0.30 | $2.50 | $731.94 | +$27 |
| Gemini 3.8 Flash (Intro Paid Tier)Google | $0.75 | $3.75 | $1,190.65 | +$486 |
| o3-mini (Reasoning)OpenAI | $1.10 | $4.40 | $1,520.95 | +$816 |
| Gemini 4 HydrogenGoogle | $1.50 | $7.50 | $2,371.77 | +$1,667 |
| GPT-4.1 (1M Long-Context)OpenAI | $2.00 | $8.00 | $2,701.84 | +$1,997 |
| GPT-5 (Flagship Reasoning)OpenAI | $1.25 | $10.00 | $2,943.2 | +$2,239 |
| Claude Sonnet 5.5 (Coding & Agents)Anthropic | $2.00 | $10.00 | $3,162.37 | +$2,458 |
| GPT-6.2 SolOpenAI | $2.00 | $10.00 | $3,162.37 | +$2,458 |
| Gemini 4 CarbonGoogle | $2.00 | $10.00 | $3,162.37 | +$2,458 |
| GPT-4o (Legacy Omni)OpenAI | $2.50 | $10.00 | $3,456.7 | +$2,752 |
| Gemini 3.1 Pro (Long-Context)Google | $2.00 | $12.00 | $3,686.42 | +$2,982 |
| Gemini 4 NeonGoogle | $3.00 | $15.00 | $4,743.55 | +$4,039 |
| Claude Opus 5.5 (Agentic Frontier)Anthropic | $4.00 | $20.00 | $6,324.73 | +$5,620 |
| Claude Fable 5.2Anthropic | $8.00 | $40.00 | $12,624.05 | +$11,919 |
| GPT-6.2 AstraOpenAI | $8.00 | $40.00 | $12,649.46 | +$11,945 |
Reasoning models spend most budget on long <think> output chains. Pre-populated at 3,000 daily requests with 3,500 input tokens, 2,800 output tokens, 30% prompt caching, and 2 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.