AI model rankings
See which models drive the most usage across AI Code Router. Rankings are based on token volume for requests routed through our API, aggregated across all customers and updated through March 2, 2026.
We currently focus on a single, up‑to‑date model from each major lab. Older generations (for example GPT‑4 or Gemini 1) are not available through the public router, so this leaderboard only shows our nine “latest generation” models.
Weekly usage of models across AI Code Router
Share of tokens routed in the last 7 days.
1
GPT‑5.2OpenAI
36%
2
Gemini 3.1 ProGoogle
22%
3
Claude 4 SonnetAnthropic
14%
4
Gemini 3.1 FlashGoogle
10%
5
GPT‑5 miniOpenAI
7%
6
DeepSeek V4DeepSeek
4%
7
Claude 4 OpusAnthropic
3%
8
Llama 4 405BMeta
2%
9
Mistral Large 2Mistral AI
2%
Top models
Token usage across AI Code Router (lifetime, all regions).
| # | Model | Provider | Tokens | Share of router |
|---|---|---|---|---|
| 1 | GPT‑5.2 Most used | OpenAI | 685.5B | 27% |
| 2 | Gemini 3.1 Pro | 542.3B | 21% | |
| 3 | Claude 4 Sonnet | Anthropic | 418.0B | 16% |
| 4 | Gemini 3.1 Flash | 312.4B | 12% | |
| 5 | GPT‑5 mini | OpenAI | 289.7B | 11% |
| 6 | DeepSeek V4 | DeepSeek | 174.2B | 7% |
| 7 | Claude 4 Opus | Anthropic | 133.6B | 5% |
| 8 | Llama 4 405B | Meta | 95.1B | 4% |
| 9 | Mistral Large 2 | Mistral AI | 73.8B | 3% |
Numbers are rounded and aggregated. For project‑level usage, visit your dashboard after signing in.