pricing

Per-token. No minimums. No surprises.

USD at launch. Local currency display (NGN, KES, ZAR, EGP, GHS, EUR) ships with the local-rails wave. Pay only for what you use — most teams spend less than $50/month.

Embeddings

Where the margin lives. Multi-lingual, multi-granularity.
Model
Input
Output
Context
BAAIbge-m3
$0.012
8K
Alibabagte-Qwen2-1.5B
$0.018
32K
MasakhaneAfroXLMR-large
$0.032
512
BAAIbge-small-en-v1.5
$0.008
512

Rerankers

Cross-encoder scoring for retrieval pipelines.
BAAIbge-reranker-v2-m3
$0.024
8K
Mixedbreadmxbai-rerank-large-v1
$0.030
512

Chat & reasoning

Small, sharp, frontier-class. We don't host 405B-vanity.
MicrosoftPhi-4
$0.18
$0.54
16K
OpenAIgpt-oss-20b
$0.22
$0.66
128K
LelapaInkubaLM-0.4B
soon
soon
2K

Speech

Per-minute audio in.
OpenAIwhisper-large-v3
$0.006/min
99 langs

All prices in USD per 1M tokens unless noted. EU egress free between Tomoul regions; standard pass-through to other clouds.

Free

$0
$5 credit · no card
  • 10 req/s rate limit
  • All public models
  • Community support
  • Shared infrastructure
Start free →
most teams

Pay-as-you-go

Per token
cards, USDC, mobile money
  • 500 req/s rate limit
  • Auto-top-up + invoicing
  • Email + Discord support
  • Multi-region routing
  • Local-currency billing
Get API key →

Enterprise

Custom
commit + dedicated
  • Dedicated capacity
  • Custom regions (Lagos, Nairobi, Cairo)
  • SOC 2 + DPA + sub-processor list
  • SLA · 99.95% uptime
  • Slack channel
Talk to sales →

Pricing FAQ

How does per-token billing work?

You pay only for the input and output tokens you actually send. We meter per request and aggregate hourly into your account balance. No minimums, no idle GPU charges, no commit.

What about local currency?

USD at launch. NGN, KES, ZAR, EGP, GHS, and EUR billing display ships with the local-rails wave (Flutterwave, Paystack, M-Pesa) alongside the Lagos POP in H2 2026. You'll be able to pay your local merchant of record, not Stripe.

Is there a free tier?

$5 free credit on signup, no card required. After that you top up with Stripe or USDC. Credits never expire.

Do you train on my data?

No. Never. By default we don't log prompts or completions, and we have no data-sharing arrangements with model providers. See /trust for residency and processor commitments.

What's the cheapest way to use Tomoul?

If your workload is privacy-sensitive or sporadic, run `tomoul serve` locally. The same CLI points at our cloud with `--cloud` when you need scale — same models, same API.