# Supported models in Hosted mode

The plugin model names you can use with a Hosted short code, the upstream provider behind each one, and the per-token price you'll be charged.

## Available plugin models

Each row is a plugin model name your client can send (in your Anthropic-compatible request body). Hosted short codes ship with these mappings pre-populated, so the first request just works. Per-token prices are what reAPI bills your wallet on Hosted requests.

| Plugin model | Upstream | Context | Input / M | Cached / M | Output / M |
| --- | --- | --- | --- | --- | --- |
| `claude-opus-4-5` (legacy) | DeepSeek V4 Pro (deepseek · deepseek-v4-pro) | 1M | $1.71/M | $0.01/M | $3.43/M |
| `claude-opus-4-6` (legacy) | DeepSeek V4 Pro (deepseek · deepseek-v4-pro) | 1M | $1.71/M | $0.01/M | $3.43/M |
| `claude-opus-4-7` | DeepSeek V4 Pro (deepseek · deepseek-v4-pro) | 1M | $1.71/M | $0.01/M | $3.43/M |
| `claude-sonnet-4-5` (legacy) | DeepSeek V4 Flash (deepseek · deepseek-v4-flash) | 1M | $0.14/M | $0.00/M | $0.29/M |
| `claude-sonnet-4-6` | DeepSeek V4 Flash (deepseek · deepseek-v4-flash) | 1M | $0.14/M | $0.00/M | $0.29/M |

## How pricing works

reAPI bills your wallet for each Hosted request. The bill = input_tokens × input_rate + cached_input_tokens × cached_rate + output_tokens × output_rate, in credits. Wallet balance ($1 = 10,000 credits, ¥1 ≈ 1,471 credits) is debited atomically after the upstream completes.

Cached input tokens are billed at the cached rate when the upstream returns a cache hit count. Anthropic-compatible upstreams (DeepSeek, Qwen) report cache_read_input_tokens on streaming responses; non-streaming responses produce no usage telemetry and are billed as if uncached.