Pricing
Pay as you go from prepaid credits. No subscription, no minimums, no daily request or token caps — you pay for what you use, per model, at the rates below.
Chat models USD per 1M tokens
| Model | Model ID | Input | Cached input | Output | |
|---|---|---|---|---|---|
| GPT-OSS 120B Popular | openai/gpt-oss-120b |
$0.20 | same as input | $0.80 | |
| DeepSeek V4 Flash Best value | deepseek-ai/DeepSeek-V4-Flash |
$0.19 | $0.04 | $0.38 | |
| DeepSeek V4.1 Flash | deepseek-ai/DeepSeek-V4.1-Flash |
$0.40 | $0.01 | $1.60 | |
| GLM 5.3 Flash | zai-org/GLM-5.3-Flash |
$0.20 | $0.04 | $0.66 | |
| Llama 3.3 70B Instruct | meta-llama/Llama-3.3-70B-Instruct-Turbo |
$0.20 | same as input | $0.50 | |
| GLM 5.3 | zai-org/GLM-5.3 |
$1.82 | $0.34 | $5.72 | |
| DeepSeek V4 Pro | deepseek-ai/DeepSeek-V4-Pro |
$1.72 | $0.17 | $5.15 | |
| MiniMax M3 | MiniMaxAI/MiniMax-M3 |
$0.39 | $0.08 | $1.56 | |
| Kimi K2.7 Code | moonshotai/Kimi-K2.7-Code |
$0.89 | $0.18 | $4.42 | |
| GPT-OSS 20B Cheapest | openai/gpt-oss-20b |
$0.04 | same as input | $0.19 | |
| Qwen3.8 Flash | Qwen/Qwen3.8-Flash |
$0.15 | $0.02 | $0.50 | |
| Qwen3-VL 235B Vision | Qwen/Qwen3-VL-235B-A22B-Instruct |
$0.26 | same as input | $1.15 |
Text to speech USD per 1M characters
| Model | Model ID | Price | |
|---|---|---|---|
| Kokoro 82M Popular | hexgrad/Kokoro-82M |
$5.20 | |
| Orpheus 3B | canopylabs/orpheus-3b-0.1-ft |
$19.50 |
Speech to text USD per audio minute
| Model | Model ID | Price | |
|---|---|---|---|
| Whisper Large v3 Popular | openai/whisper-large-v3 |
$0.002 | |
| Parakeet TDT 0.6B v3 Fastest | nvidia/parakeet-tdt-0.6b-v3 |
$0.002 | |
| Voxtral Small 24B New | mistralai/Voxtral-Small-24B-2507 |
$0.004 | |
| Voxtral Mini 3B Cheapest | mistralai/Voxtral-Mini-3B-2507 |
$0.0013 |
Embeddings USD per 1M input tokens
| Model | Model ID | Price | |
|---|---|---|---|
| BGE-M3 Popular | BAAI/bge-m3 |
$0.02 | |
| Qwen3 Embedding 8B | Qwen/Qwen3-Embedding-8B |
$0.02 |
Rerank USD per 1M input tokens
| Model | Model ID | Price | |
|---|---|---|---|
| Qwen3 Reranker 8B | Qwen/Qwen3-Reranker-8B |
$0.07 |
How billing works
Add funds with a card; each request deducts its exact cost from your balance. You can test every model.
Chat is billed on prompt + completion tokens (reasoning tokens included); prompt tokens the backend serves from its cache bill at the lower cached-input rate, automatically, speech on input characters, transcription on audio minutes measured from your file.
There is no free-tier ceiling to outgrow and no upgrade to be approved for. Requests and tokens are not metered per minute or per day; only abusive traffic is throttled.
Prices are in USD and may change as upstream costs change; the model page always shows the current rate. Questions? Contact us.