Perks

Claude Startups members get offers worth up to $45,000 from companies building with Claude, biweekly group sessions with Anthropic's Applied AI team, and invites to Founder Days, Builder Days and hackathons. The current cohort is full and new applications are paused. Startups backed by a VC in Anthropic's partner network can claim up to $100,000 in API credits through that VC.

OpenAI distributes API credits through partner channels. VC portfolios, accelerators, Ramp, Brex and the Grove program, from $2,500 for early companies to larger grants for backed startups, plus events and early access.

Mistral's startup program (Mistralship) provides La Plateforme credits, technical support and events for early companies building on Mistral's open-weight and frontier models, with European data hosting.

Together AI's startup accelerator gives AI-native startups inference and fine-tuning credits on open models (Llama, DeepSeek, Qwen, FLUX) plus engineering support.

ElevenLabs Startup Grants give early companies 12 months of access with 33 million characters, over 680 hours of Agents usage, for text-to-speech, voice agents and dubbing. ElevenLabs values the grant at more than $5,500.

Deepgram's startup program funds speech-to-text, text-to-speech and voice-agent usage for new customers building voice products.

AssemblyAI offers early-stage startups transcription and audio-intelligence credits plus technical support for products built on voice data.

Runway's Builders Program supports startups building on Runway's video models with API credits, early access and introductions to Runway's venture fund.

Baseten runs dedicated inference and training for AI-first companies. The startup program gives up to $25,000 in credits for dedicated inference or training deployments plus up to $2,500 for Baseten Model APIs, with support responses within a few business hours, cross-cloud autoscaling and launch amplification.

Fireworks serves open models (Llama, DeepSeek, Qwen, Mistral, FLUX, Whisper) at production speed. The startup program grants credits, sized per company, valid for one year, that cover text and vision inference, image generation, speech-to-text, embeddings, fine-tuning and on-demand GPU deployments, plus raised rate limits.

Runware is one inference API for image, video, audio, LLM and 3D models. The Starter tier gives bootstrapped and pre-seed teams up to $1,000 in credits with no deposit, awarded in $100 steps based on your application. The Growth tier adds $10,000 on a $25,000 deposit and the Scale tier adds $30,000 on a $50,000 deposit. Credits do not expire and work across every model and modality.