FlashAPI
← Back to FlashAPI

FlashAPI research tool · public prices

AI model token cost calculator for Southeast Asia

Enter your expected monthly requests and average input and output tokens per request. The calculator applies each provider’s listed API rate to that same workload, so you can prepare a shortlist before running quality and latency tests.

Your monthly workload

Inputs stay in this browser. This calculator does not submit your workload to FlashAPI or write it to D1.

Calculated token volume

Monthly input tokens
100,000,000
Monthly output tokens
30,000,000

Public-rate estimates

Compare only within the same currency and rate condition. These use published provider rates for the stated conditions, not FlashAPI prices or confirmed availability. A dash means the request exceeds this row’s shown tier or model limits.

USD published rates

USD published rates
Provider / modelInput rateOutput rateEstimated monthly token billRate condition
XiaomiMiMo-V2.6-FlashOfficial source ↗Checked: 2026-09-27$0.14per 1M tokens$0.28per 1M tokens$22.40Overseas real time · serving region unverified
Z.AIGLM-5.3-FlashOfficial source ↗Checked: 2026-09-27$0.15per 1M tokens$0.50per 1M tokens$30.00Published USD rate · serving region unverified
DeepSeekV4.1-FlashOfficial source ↗Checked: 2026-09-27$0.15per 1M tokens$0.60per 1M tokensOff-peak: $33.00Peak: $66.00Off peak; peak $0.30/$1.20 · region unverified
MiniMaxMiniMax-M3Official source ↗Checked: 2026-09-27$0.30per 1M tokens$1.20per 1M tokens$66.00Standard ≤512K input tier · published 50% discount · region unverified

CNY published rates

CNY published rates
Provider / modelInput rateOutput rateEstimated monthly token billRate condition
Alibaba CloudQwen3.8-FlashOfficial source ↗Checked: 2026-09-27CN¥1.094per 1M tokensCN¥3.427per 1M tokensCN¥212.21Singapore international · real time · text
Tencent CloudHy3Official source ↗Checked: 2026-09-27CN¥1.00per 1M tokensCN¥4.00per 1M tokensCN¥220.00Singapore · online inference · text
VolcengineDoubao-Seed-2.1-LiteOfficial source ↗Checked: 2026-09-27CN¥0.80per 1M tokensCN¥2.70per 1M tokensCN¥161.00CNY online · non-audio input ≤1M · SEA region unverified
BaiduERNIE-4.5-Turbo-128KOfficial source ↗Checked: 2026-09-27CN¥0.80per 1M tokensCN¥3.20per 1M tokensCN¥176.00CNY online · 128K version · SEA region unverified

How the estimate is calculated

Monthly cost = (requests × average input tokens ÷ 1,000,000 × input rate) + (requests × average output tokens ÷ 1,000,000 × output rate)At the starting values of 100,000 requests, 1,000 input tokens and 300 output tokens, the workload is 100 million input and 30 million output tokens. Change any field above to recalculate the table instantly.

Read this before choosing a model

  • USD and CNY are shown separately. No exchange rate or tax is applied. A cheaper number in one currency is not a cross-currency ranking.
  • The listed row conditions matter. DeepSeek’s displayed rate is off-peak; its peak rate differs. Some prices refer to China or international billing without a verified Southeast Asian serving region.
  • The estimate excludes cached-input discounts, tools or search fees, reasoning-token differences, retries, minimum commitments, other promotions and contract discounts. Check the linked official price page before buying.
  • Token price does not predict answer quality, data location or latency. Test your own Indonesian, Vietnamese and Thai tasks from your users’ countries.
  • FlashAPI has no connected upstream provider or live model access yet. These are research candidates, not models offered through FlashAPI.

Frequently asked questions

Does this show the price I would pay FlashAPI?

No. The calculator uses providers’ published public rates for specific conditions to size a potential workload. FlashAPI has not launched model access or published resale pricing.

Why are input and output tokens separate?

Providers often charge different rates for tokens sent in a prompt and tokens generated in the answer. An application that produces long answers can cost much more than an input-only comparison suggests.

Can I use the lowest result as my model choice?

Treat it as a candidate for testing. Confirm access, region, data terms, quality and p95 first-token and full-response time on representative traffic before choosing a route.

Continue the evaluation

Our selection guide explains the cost example and a repeatable Southeast Asian latency test. The API documentation shows the planned unified gateway contract and current early-access status.