> Heroku AI

www.heroku.com

Heroku AI is a managed inference and agents layer on top of the Heroku platform, billed per 1M tokens with the model, input and output rates published directly on the Heroku pricing page. Examples include claude-opus-4-8 at $5.50 input / $27.50 output, claude-4-5-sonnet at $3.30 / $16.50, gpt-oss-120b at $0.15 / $0.60, qwen3-235b at $0.22 / $0.88, glm-4-7-flash at $0.07 / $0.40 and nova-lite at $0.06 / $0.24, with each model tagged by which region it is provisioned in (US, EU, or both). MCP servers run as one-off dynos on a pay-as-you-go basis, and the platform retains its dyno grid from Eco at $5/month up to Shield-2XL at $1,800/month.

unknownbest-effort probe · checked never

recent reviewsedit

Paid API API keyPaid-onlyLow risk· conf 90%verified

Free credits

No free tier

No free inference credits or AI allowance are stated on the Heroku pricing page; the only credit mechanism described is Enterprise add-on credits where one credit equals $1 of marketplace services (checked 2026-09-16).

Cheapest input /1M

Free

USD per 1M input tokens, best tracked model

Cheapest output /1M

Free

USD per 1M output tokens, best tracked model

Models tracked

Canonical models with normalized pricing

// source_analysisUprouter editorial analysis

// pros

Per-model list prices for input and output are published on the pricing page itself, including regional provisioning (US and/or EU) for each model
Multi-vendor catalog on one billing relationship covers Anthropic, Amazon Nova, OpenAI gpt-oss, DeepSeek, Qwen, MiniMax, Moonshot and Z.ai models
MCP Toolkits let agentic clients such as Claude Desktop, Cursor and Heroku Managed Inference call multiple MCP servers through one gateway, with those servers billed as pay-as-you-go one-off dynos
pgvector on Heroku Postgres keeps embeddings and similarity search inside the same trusted database rather than a separate vector service
Compliance path is explicit at the dyno layer: Shield dynos allow HIPAA BAAs, PCI compliance and strict TLS enforcement, and Private Spaces give full network isolation

// cons

No free AI allowance or credit is published: the AI section lists per-token rates only, and add-on credits (1 credit = $1) are described for Enterprise customers in the marketplace rather than as an AI grant
Platform cost sits on top of token cost, since inference runs alongside dynos that start at $5/month (Eco, which sleeps after 30 minutes) and rise steeply for compliance-grade Shield tiers
Only a subset of models is provisioned in both US and EU, so EU data-locality requirements narrow the usable catalog
Model coverage is narrower than a frontier-lab direct API, with no first-party frontier models from Anthropic's newest line beyond the listed Claude variants and no Google or Meta Gemini-class frontier models in the published table

A practical pick for teams already shipping on Heroku that want managed inference and MCP tooling without a second vendor relationship, especially where pgvector and the existing pipeline/review-app workflow matter. Where EU residency or HIPAA is required, budget for Private or Shield dynos plus the EU-provisioned model list rather than the cheapest tokens.

written by Mike Fleming, Uprouter editorial · verified against official sources September 16, 2026

// free_quota

No free tier

This provider does not offer a free tier as of our last verification (Sep 16, 2026). The cheapest way to try it is pay-as-you-go — input from Free per 1M tokens. If the provider adds a free tier, submit a correction and we’ll re-verify.

// pricinglast verified Sep 16, 2026

Plan	Type	Monthly	Input /1M	Output /1M	Note
Pay-as-you-go	payg	—	Free	Free	Cheapest tracked model rate (OpenRouter-normalized).

pricing_last_verified: Sep 16, 2026official pricing page

Pricing normalized from public sources — always verify with the provider.

// model prices

Model	Input /1M	Output /1M	Context
Nova Lite 1.0	Free	Free	300k
MiniMax M2	Free	$0.0000	205k
Qwen: Qwen3 72B	Free	Free	131k

// faq

$ Does Heroku AI have a free tier?

No. Heroku AI does not offer a meaningful free tier — entry requires payment (pay-as-you-go or a plan). We mark this explicitly rather than showing a misleading zero.

$ How much does Heroku AI cost per 1M tokens?

The cheapest model we track at Heroku AI is Free per 1M input tokens and Free per 1M output tokens. This is normalized from Heroku AI's published pricing — verify with the provider before purchasing, since prices change frequently.

$ Is Heroku AI safe to use?

Uprouter rates Heroku AI as low risk. Risk level: moderate. Billing is transparent per 1M tokens with no commitment or minimum-spend clause visible, and compliance options (HIPAA BAA, PCI, network isolation, TLS enforcement) are documented at the Shield and Private dyno tiers, which lowers regulatory risk for regulated apps. The offsetting risks are purely cost-side: no published free AI allowance, and the effective bill is token rate plus always-on dyno cost, which can dominate the inference line for low-traffic apps. Risk ratings are editorial, evidence-based, and never influenced by affiliate relationships.

// community0 reviews

Raw community sentiment, shown unblended. Votes never feed into objective data fields like price, risk, or status — and never into sort order.

No published reviews yet — be the first.

// more_informationprice history · live status · risk & confidence · code snippets

price_history

No recorded changes yet.

live_status

current: unknown

No probes recorded yet.

Uprouter runs its own probes; best-effort, not a guarantee. Probe results depend on our network location and cadence — for production, always use the provider’s own status page.

risk_&_confidence

Low risk

90%

Risk level: moderate. Billing is transparent per 1M tokens with no commitment or minimum-spend clause visible, and compliance options (HIPAA BAA, PCI, network isolation, TLS enforcement) are documented at the Shield and Private dyno tiers, which lowers regulatory risk for regulated apps. The offsetting risks are purely cost-side: no published free AI allowance, and the effective bill is token rate plus always-on dyno cost, which can dominate the inference line for low-traffic apps.

Confidence reflects how much verifiable evidence (official docs, probe history, community reports) backs this entry. Risk is our editorial assessment of operator reliability and terms — not financial advice.

code_snippets

# Heroku AI's own API — get a key from their dashboard
curl https://www.heroku.com/chat/completions \
  -H "Authorization: Bearer THEIR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "amazon/nova-lite-v1",
    "messages": [{ "role": "user", "content": "Hello" }]
  }'

Direct to the provider — Uprouter receives nothing; verify their current pricing before purchasing.

// similar_providersscored by Paid API signals

Ranked by neutral similarity signals only — live status, free-tier shape and free-credit proximity. Commercial relationships never influence this list.