DeepSeek V4 Flash
$0.14 in · $0.28 out · cache $0.003
Wildly cheap. With cached input it's basically free. Western alternative: Gemini 2.5 Flash-Lite ($0.10 / $0.40).
Prices update continuously from the providers' own pages plus the OpenRouter live feed. Anything that drifts more than 20% from our hand-checked seed gets a flag — we don't trust automated data more than we trust ourselves.
$0.14 in · $0.28 out · cache $0.003
Wildly cheap. With cached input it's basically free. Western alternative: Gemini 2.5 Flash-Lite ($0.10 / $0.40).
$2-3 in · $8-15 out
GPT-4.1 wins on price + 1M context. Sonnet wins on instruction-following and aggressive caching (cache read $0.30).
$1.10 in · $4.40 out
Same price as o3-mini, newer generation. For code and math it punches above its price class.
$0.02 in
The cheapest serious embeddings model. Use 3-large ($0.13) only if your semantic search actually fails at small.
Honest take: prices below are correct, but the "best model" changes monthly. This shortlist updates with the same cron that updates the table.
| Provider | Model | Input $/1M | Output $/1M | Cached $/1M | Context | Type | Try |
|---|---|---|---|---|---|---|---|
| Loading prices… | |||||||
Affiliate disclosure: "Try" buttons link to each provider's own signup page. When we have an affiliate code we'll mark the button "Try (aff)" and use a sponsored rel attribute — never silently. Right now no link is affiliated. Pricing is identical for you either way.
Every price in this table starts from data/pricing.json in our repo, checked against each provider's official pricing page.
OpenRouter aggregates pricing from ~300 models with community updates. Our endpoint pulls and merges every 60 minutes.
If the live price for a model differs from our seed by more than 20%, we keep the seed and flag it. Catches OpenRouter glitches and unannounced provider changes.
Hit GET /api/pricing and you get the same data we use, with a 10-minute CDN cache. Build your own dashboard. Be our guest.