Pricing
Pay for what you use.
Pay-as-you-go credit in THB. No monthly commitments, no foreign card needed. Receive a Thai tax invoice with every top-up, and a full tax invoice on request.
How billing works
Top up once. Spend across any model.
- 01
Top up in THB
PromptPay from a Thai bank account. Minimum ฿300 before 7% VAT — below that, fees eat the top-up.
- 02
Spend across the catalogue
One balance covers every model. Input, output and cached tokens are priced separately, because providers charge for them separately.
- 03
Rate locked at request time
Each usage record keeps the rate that applied when the request happened. A later price change never re-prices what you already paid for.
The formula
How a THB rate is built.
We pin the FX rate rather than float it with the market, then add our margin, which is set per model. Current per-model rates are in the catalogue on this page — each request is billed at the rate locked when it ran.
Minimum top-up ฿300 before 7% VAT. Unused credit does not expire.
Included
The same deal for every customer.
- One published rate per model — provider cost plus our margin, never negotiated per account.
- Billed in THB via PromptPay — no foreign card needed.
- Rate locked when the request happens — no retroactive re-pricing.
- Unused credit never expires on a timer.
- A Thai tax invoice on every top-up, with a full tax invoice on request.
- FX rate pinned, not floating with the market.
Catalogue
Current per-model rates.
THB per million tokens. You call a model by name and the gateway chooses where it runs according to your organisation’s routing policy, so you never pick a provider yourself. Sort by price, or rank by strength (coding, speed, value, and more). Rates can change; a request never gets re-priced after it runs.
Scores are a guide for comparison, not live benchmarks.
| Model | Highlights | Input / 1M tokens | Output / 1M tokens | Cache read / 1M tokens | Cache write / 1M tokens |
|---|---|---|---|---|---|
Text Embedding 3 Small openai/text-embedding-3-small | OpenAI 1,536-dim embeddings — cheapest API tier; MTEB 62.3 at scale. | ฿0.96 | ฿0.00 | — | — |
Nemotron Lightning 3.5 30B fw-nemotron-lightning-3.5-30b | Nemotron Lightning 3.5 30B — NVIDIA hybrid MoE (3B active); MMLU-Pro 81.9, 1M context. | ฿1.92 | ฿7.69 | ฿0.38 | — |
Qwen3 Embedding 8B fw-qwen3-embedding-8b | Qwen3 Embedding 8B — MTEB 70.6 leader; top open-weight retrieval on Fireworks. | ฿3.85 | ฿0.00 | — | — |
Embed 4 cohere/embed-4 | Cohere Embed v4 — MTEB 65.2; multilingual text & image vectors for RAG. | ฿4.62 | ฿0.00 | — | — |
GLM-5.3 Flash glm-5.3-flash | GLM-5.3 Flash — Terminal-Bench 2.1 84.3%; first native-multimodal GLM-5, 18B active. | ฿5.77 | ฿19.23 | ฿1.15 | — |
GPT-OSS 120B gpt-oss-120b | GPT-OSS 120B — OpenAI open weights; SWE-Bench Verified 62.4%, near o4-mini. | ฿5.77 | ฿23.08 | ฿0.58 | — |
Qwen3.8 Flash qwen/qwen3.8-flash | Qwen3.8 Flash — SWE-bench Pro 62.5%; fast multimodal agent coding on DashScope. | ฿5.77 | ฿18.08 | — | — |
Text Embedding 3 Large openai/text-embedding-3-large | OpenAI 3,072-dim embeddings — MTEB 64.6; strong default for high-quality RAG. | ฿6.00 | ฿0.00 | — | — |
GPT-5.6 Luna openai/gpt-5.6-luna | GPT-5.6 Luna — OpenAI’s fastest, cheapest tier for high-volume chat & agents. | ฿7.69 | ฿46.16 | ฿0.77 | ฿9.62 |
GPT-5 Mini openai/gpt-5-mini | GPT-5 Mini — 400K context; fast, cost-efficient GPT-5 for high-volume tasks. | ฿9.62 | ฿76.94 | ฿0.96 | — |
Rates may change. You pay the rate that applied when the request ran.
Honest about the price
Above provider list, on purpose.
Our rates are usually a little above contracting a foreign cloud directly. The difference pays for what we handle locally: THB billing, Thai tax invoices, data stored on AWS in Thailand, and support during Thai business hours.
See it on a real request.
Generate a key in your dashboard, copy the sample code from our docs, and send your first API request instantly.