Stop guessing what your AI API bill should be.
The same open model is often sold at 3–8 different prices across providers. APIRadar tracks official per-token prices from 9 major providers, estimates your Claude Code / Codex CLI / Cline monthly spend, and shows you the exact config to switch to a cheaper endpoint.
Browse the price database → Estimate my monthly cost Wire a cheap model into my agent
What you can do here
📊 Compare providers
Which provider sells the same model family cheapest — including cache-read pricing and context window. Every number links to the official source page.
Open /pricing →🧮 Estimate monthly cost
Enter your token usage, get a monthly bill for every provider selling that model, plus the top-3 cheapest with deep links.
Open /calculator →🔧 Switch your agent
Copy-paste configs pointing Claude Code, Codex CLI and Cline at DeepSeek, GLM (Z.ai) or SiliconFlow — real env vars from official docs.
Open /configs →TL;DR from the 2026-09-17 sweep
- Cheapest hosted DeepSeek-class model we verified: DeepSeek-V4-Flash on SiliconFlow — $0.13 in / $0.28 out per 1M (official DeepSeek peak is $0.30/$1.20).
- Free tiers exist: Z.ai lists GLM-4.7-Flash and GLM-4.5-Flash at $0.
- Aggregators don't always cost more: OpenRouter sells Qwen3.7-Plus ($0.32/$1.28) under Alibaba's own list price ($0.40/$1.60).
- Claude-compatible endpoints are everywhere now: DeepSeek and Z.ai both expose Anthropic-format base URLs, so Claude Code can point at them by changing one env var.