Providers
TOKEN owns no GPUs. Each request goes to one of these independent providers, chosen by the routing rules; only that provider receives the request's content. This page is generated from the same catalog that sets routing and prices, last updated October 24, 2026.
"Data location" is where the provider says it processes requests, from its own published terms. Providers marked US/EU are used by the us-eu-hosted group. Precision is as each provider states it; TOKEN routes to providers serving the maker's native weights first.
DeepInfra
Data location: US · in the us-eu-hosted group
| Model | Stated precision | Checked |
|---|---|---|
| BGE-M3 | FP32 | October 19, 2026 |
| DeepSeek V4.1 Flash | FP8 native | October 19, 2026 |
| gpt-oss-120b | BF16 | October 19, 2026 |
| gpt-oss-20b | BF16 | October 19, 2026 |
| Kimi K3 | MXFP4 native | October 19, 2026 |
Fireworks AI
Data location: Not stated by the provider · not in the us-eu-hosted group
| Model | Stated precision | Checked |
|---|---|---|
| GLM-5.3 | FP8 native | October 19, 2026 |
| GLM-5.3 Flash | Not stated | October 19, 2026 |
| Kimi K3 | MXFP4 native | October 19, 2026 |
Novita AI
Data location: US · in the us-eu-hosted group
| Model | Stated precision | Checked |
|---|---|---|
| BGE-M3 | Not stated | October 19, 2026 |
| DeepSeek V4 Pro | FP8 native | October 19, 2026 |
| Gemma 4 26B A4B | BF16 | October 19, 2026 |
| Gemma 4 31B | BF16 | October 19, 2026 |
| GLM-5.3 | FP8 native | October 19, 2026 |
| GLM-5.3 Flash | FP8 native | October 19, 2026 |
| Kimi K3 | Not stated | October 19, 2026 |
| Llama 4 Maverick | FP8 native | October 19, 2026 |
Z.ai
Data location: SG · not in the us-eu-hosted group
| Model | Stated precision | Checked |
|---|---|---|
| GLM-5.3 | NATIVE native | October 19, 2026 |
| GLM-5.3 Flash | NATIVE native | October 19, 2026 |