Use-case preset
Hobby project (1M tokens/mo) cost calculator
Smallest, cheapest models for personal projects on a 1M token/month budget.
Building something for fun or a side project at a 1M tokens/month budget. At $0.06–0.20 per million input tokens for the smallest models, 1M tokens buys you 5,000–16,000 typical 4k-context interactions per month — plenty for personal use or a small beta. The 70/30 input/output ratio and 4k context reflect general-purpose conversational or generation tasks.
This preset is optimized for cost, not capability. Llama 3.2 1B and 3B, Gemma 2 2B, Phi-3 Mini, and Qwen 3 8B all run under $0.10/M input tokens through the cheapest providers and handle surprisingly wide task ranges. fp8 and int4 quantized variants cut cost further with minimal quality loss for casual applications. Skip caching configuration at this volume — the overhead is not worth the engineering time. Track your monthly token burn with a simple counter; the hobby budget evaporates faster than expected when running multi-turn conversations.