OpenRouter Model Token Pricing Guide | Generated by AI
Here’s what those exact OpenRouter models currently cost per 1M tokens (input → output). I’m pulling straight from each model’s page:
- Anthropic Claude Opus 4 — $15 → $75 (images: $24/K). (OpenRouter)
- Anthropic Claude Sonnet 4 — $3 → $15 (images: $4.80/K). (OpenRouter)
- Google Gemini 2.5 Flash — $0.30 → $2.50 (images: $1.238/K). (OpenRouter)
- Google Gemini 2.5 Pro — $1.25 → $10 (page lists “starting at”). (OpenRouter)
- DeepSeek V3 — $0.18 → $0.72. (OpenRouter)
- Moonshot Kimi K2 — $0.14 → $2.49. (OpenRouter)
- Mistral Medium 3.1 — $0.40 → $2.00. (OpenRouter)
- Qwen3 Coder — $0.20 → $0.80 (note: higher tier kicks in past 128k input tokens). (OpenRouter)
- OpenAI GPT-OSS 120B — $0.073 → $0.29. (OpenRouter)
- OpenAI GPT-5 — $1.25 → $10. (OpenRouter)
Quick tips & gotchas:
- Your mapping has two
deepseek-v3keys; in Python dicts the later one overwrites the earlier. If you want both paid and:freeroutes, use distinct keys (e.g.,"deepseek-v3-free"vs"deepseek-v3"). - Some models have
:freeroutes (useful for quick tests but rate-limited): e.g.,qwen/qwen3-coder:freeandmoonshotai/kimi-k2:free. (OpenRouter)
If you want, I can compute example request costs for your typical prompt/response token sizes.