AI models & APIs credits for startups

Three tiers. Start gives $2,000 self-serve. Scale covers 100% of usage up to $100K, then 20% up to another $100K. AI-First adds $150K on top, reaching $350K for AI-core teams.

The largest ceiling in the database. A small self-serve tier needs just a live product; six-figure tiers route through Activate Providers — accelerators, incubators, VCs. Credits cover third-party models via Bedrock.

Works across Azure, GitHub and other Microsoft platforms. Entry tier is self-serve, no funding needed — a real bootstrapped on-ramp. Top tiers need VC backing through Microsoft's investor network.

Free Claude API credits plus priority rate limits so you can run production traffic without throttling. Anthropic does not publish the credit amount. Joining the program is open, but the credits specifically require institutional equity funding. Members get early notice of model releases and access to Builder Days, hackathons and Founder Days.

A five-week cohort program open to founders at any stage, including pre-idea and pre-product. Participants receive $50,000 in API credits, early access to unreleased tooling and hands-on mentorship from OpenAI staff. This is the largest OpenAI credit allocation a founder can reach without a VC or accelerator relationship, and it is competitive rather than automatic.

OpenAI does not run a broad self-serve credit program. Allocations are routed through partners: a business banking or spend-management relationship commonly unlocks around $2,500, while VC and accelerator partners can nominate startups for more. Direct applications without a partner relationship are rarely accepted, so the practical route is through whoever already banks or backs you.

Up to €30,000 in La Plateforme API credits for selected early-stage startups, plus one-to-one support from Mistral's Solutions and Science teams. Worth applying if any part of your stack can run on open-weight European models — the review is selective but the application does not require a VC introduction.

One of the more generous open-model inference programs, with the allocation scaling by which partner refers you. Useful if you are serving Llama, Qwen, DeepSeek or other open weights and want to move off per-token frontier pricing without standing up your own GPUs.

Twelve months of the full ElevenLabs platform — Conversational AI, text-to-speech, speech-to-text, voice cloning, sound design and music — with 33 million characters included, roughly 680 hours of agent usage. Rolling applications, decisions within about a week. Note the branding condition: recipients display the ElevenLabs Grants logo on their site for at least 12 months.

Sonar API credits plus Enterprise Pro seats for startups coming in through an approved accelerator or VC partner. Narrow but genuinely useful if your product needs search with grounded, sourced citations rather than raw generation.