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.
Simulate sending full codebases and telemetry logs into Gemini Pro with context caching.
Multi-turn history expands average input context to 61,580 tokens/req. Output accounts for 33% of total spend.
| Model | Delta | |||
|---|---|---|---|---|
| Claude Haiku 5.5 (High-Volume)Anthropic | $0.10 | $0.50 | $44.27 | Save $892 |
| DeepSeek V4.1 Flash (Off-Peak)DeepSeek | $0.15 | $0.60 | $51.78 | Save $885 |
| GPT-4.1 nano (Ultra-Low Cost)OpenAI | $0.10 | $0.40 | $55.21 | Save $881 |
| GPT-4o mini (Legacy Budget)OpenAI | $0.15 | $0.60 | $116.55 | Save $820 |
| Gemini 3.5 Flash-Lite (Budget Edge)Google | $0.30 | $2.50 | $158.39 | Save $778 |
| DeepSeek V4 Pro (Off-Peak Reasoning)DeepSeek | $0.66 | $1.98 | $218.86 | Save $718 |
| GPT-4.1 mini (Balanced Fast)OpenAI | $0.40 | $1.60 | $220.84 | Save $716 |
| Gemini 3.8 Flash (Intro Paid Tier)Google | $0.75 | $3.75 | $332.05 | Save $605 |
| Gemini 4 HydrogenGoogle | $1.50 | $7.50 | $596.61 | Save $340 |
| GPT-5 (Flagship Reasoning)OpenAI | $1.25 | $10.00 | $649.29 | Save $287 |
| Llama 3.3 / 4 70B-Class (Hosted API)Open-Weight Host | $0.59 | $0.79 | $683.72 | Save $253 |
| Claude Sonnet 5.5 (Coding & Agents)Anthropic | $2.00 | $10.00 | $795.49 | Save $141 |
| GPT-6.2 SolOpenAI | $2.00 | $10.00 | $795.49 | Save $141 |
| Gemini 4 CarbonGoogle | $2.00 | $10.00 | $795.49 | Save $141 |
| o3-mini (Reasoning)OpenAI | $1.10 | $4.40 | $854.74 | Save $82 |
| Gemini 3.1 Pro (Long-Context)Google | $2.00 | $12.00 | $936.59 | Active Baseline |
| GPT-4.1 (1M Long-Context)OpenAI | $2.00 | $8.00 | $1,104.22 | +$168 |
| Gemini 4 NeonGoogle | $3.00 | $15.00 | $1,193.23 | +$257 |
| Claude Opus 5.5 (Agentic Frontier)Anthropic | $4.00 | $20.00 | $1,590.97 | +$654 |
| GPT-4o (Legacy Omni)OpenAI | $2.50 | $10.00 | $1,942.58 | +$1,006 |
| Claude Fable 5.2Anthropic | $8.00 | $40.00 | $3,002 | +$2,065 |
| GPT-6.2 AstraOpenAI | $8.00 | $40.00 | $3,181.94 | +$2,245 |
Context caching turns massive repo reads into an affordable engineering review loop. Pre-populated at 600 daily requests with 60,000 input tokens, 1,400 output tokens, 80% prompt caching, and 3 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.