Find the best AI model route for your app
Compare 271 indexed models by price, latency, uptime, throughput, context window, and capabilities — all in one developer-friendly workspace.
Long-context examples
Tracked models
10
23 endpoint rows
Chart models
10
3 snapshot hours
Models updated
May 17, 7:37 AM
Fresh Supabase snapshot
Endpoint updated
May 17, 7:37 AM
Newest snapshot May 17, 7:37 AM
Built around model decisions developers actually make
Lowest latency route
101ms
Llama 3.3 70B Instruct via groq
Best uptime route
100%
Llama 3.3 70B Instruct via groq
Top throughput route
278.5 tok/s
Gemini 3 Pro Preview via vertex
Endpoint snapshot
May 17, 7:37 AM
23 provider routes tracked
Route recommendations
Best route in the latest AI Gateway snapshot
Claude Sonnet 4.5
via vertexAnthropic
latency
637ms
uptime
100%
price
$18.00 / 1M
GPT 5.2
via azure
latency
443ms
uptime
100%
price
$15.75 / 1M
GPT 5.1 Thinking
via openai
latency
637ms
uptime
100%
price
$11.25 / 1M
Gemini 3 Pro Preview
via google
latency
2.78s
uptime
100%
price
$14.00 / 1M
Balanced score uses 45% latency rank, 35% uptime rank, and 20% price rank across tracked endpoints for each model.
| Model | Provider | Latency | Uptime | Throughput | Input | Output | Signal |
|---|---|---|---|---|---|---|---|
| Llama 3.3 70B Instruct | groq | 101ms | 100% | — | $0.59 / 1M | $0.79 / 1M | Lowest latency |
| Llama 3.3 70B Instruct | bedrock | 157ms | 100% | 189 tok/s | $0.72 / 1M | $0.72 / 1M | Tracked route |
| Mistral Large 3 | mistral | 371ms | 100% | 61 tok/s | $0.50 / 1M | $1.50 / 1M | Tracked route |
| GPT 5.2 | azure | 443ms | 100% | — | $1.75 / 1M | $14.00 / 1M | Tracked route |
| DeepSeek V3.2 | bedrock | 556ms | 100% | 64 tok/s | $0.62 / 1M | $1.85 / 1M | Tracked route |
| GPT 5.1 Thinking | openai | 637ms | 100% | 140.5 tok/s | $1.25 / 1M | $10.00 / 1M | Tracked route |
| Claude Sonnet 4.5 | vertexAnthropic | 637ms | 100% | 48 tok/s | $3.00 / 1M | $15.00 / 1M | Tracked route |
| DeepSeek V3.2 | deepseek | 652ms | 100% | 83 tok/s | $0.28 / 1M | $0.42 / 1M | Tracked route |
Endpoint metrics are read from Supabase AI Gateway snapshots. Pricing and availability may change.
Model choice is now a production decision
The cheapest model is not always the cheapest route once latency, fallbacks, reliability, and output quality are included.
Compare model pricing without jumping across provider pages
Track endpoint latency and uptime over time
Estimate cost before shipping an AI feature
Choose better fallback routes for production apps
Built on snapshots, ready for richer routing
GitHub Actions captures AI Gateway data into Supabase snapshots. Historical runs can power latency, uptime, throughput, pricing trends, and route recommendations as the dataset grows.
Vercel AI Gateway API
Snapshot storage
Trend charts
Route recommendation
Model directory
A structured index of popular AI models with pricing, windows, modalities, and provider availability.
Endpoint comparison pages
Dedicated comparison pages for model routes across providers, regions, and gateway endpoints.
AI cost calculator
A usage calculator for estimating monthly API spend before a feature ships to production.
Historical trend charts
Latency, uptime, throughput, and pricing snapshots over time for better routing decisions.