Free AI API tiers & signup credits

New Google Cloud customers get a $300 Welcome credit to spend over a 90-day Free Trial on most Google Cloud services. You must add a valid payment method, but you are not billed during the trial. The credit cannot pay for the Gemini API in AI Studio or for partner models offered as managed APIs, and trial accounts cannot add GPUs, use Marketplace or request quota increases. If you don't upgrade, resources are stopped at the end and deleted after a 30-day grace period. Separately, the always-free Free Tier keeps running with no end date.

Every new Deepgram account gets $200 in free credit on the Pay As You Go plan, with no credit card required. Deepgram says this is about 43,000 minutes (over 700 hours) of Nova transcription, and the credit works across speech-to-text, text-to-speech, the Voice Agent API and audio intelligence. Deepgram's pricing page says the credit lasts until used, but its developer docs say free promotional credits expire one year from signup.

AssemblyAI gives new accounts $50 in free credits with no credit card required. AssemblyAI's pricing FAQ says this covers up to 185 hours of pre-recorded or 333 hours of streaming transcription, depending on the model. Free accounts can use pre-recorded transcription, Speech Understanding and streaming, but not the LLM Gateway, and have lower concurrency limits until they upgrade.

Google's Gemini API has a Free tier where many models are listed as 'Free of charge' for input and output, including Gemini 3.8 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini Embedding 2 and Gemma 4. Newer flagship, image, video and music models (Gemini 4 Argon, Gemini 3.1 Pro Preview, Nano Banana, Veo 3.1, Lyria) are paid-only. Free-tier prompts and responses may be used to improve Google's products and can be read by human reviewers. Rate limits apply per project and are shown in Google AI Studio rather than published as a fixed table.

GroqCloud has a Free plan with per-model rate limits published on its rate-limits docs page, for models including gpt-oss-120b, gpt-oss-20b, Qwen3.8-27B, Whisper and Orpheus text-to-speech. The paid Developer plan is pay-as-you-go with much higher limits (for example 1K RPM instead of 30 for gpt-oss-120b) plus Batch and Flex processing. Limits apply at the organization level and your exact numbers are shown on the console Limits page.

Mistral's docs state that API access is enabled by default in Free mode with no credit card required, with usage and rate limits applied. Free mode is the default state for new accounts: you can create API keys and use included monthly usage within the limits shown on your organization's Limits page. Mistral used to call this the 'Experiment' plan; the current docs call it Free mode and treat pay-as-you-go as a prepaid billing option rather than a separate plan.

OpenRouter offers ':free' variants of selected models that cost nothing per token, plus a Free Models Router (openrouter/free) that picks a free model for you. Free-model usage is limited to 20 requests per minute and 50 requests per day, rising to 1,000 requests per day once you have purchased at least 10 credits (all time). As of 2026-10-17 the public models API listed 16 free variants, including Gemma 4, NVIDIA Nemotron 3 and 3.5, Thinking Machines Inkling and Poolside Laguna models.

New Jina AI users get an auto-generated API key with free tokens usable across any Jina model; Jina's Reader page states that every new API key comes with 10M free tokens. Keys don't expire, no card is needed, and the Reader API also works without a key at 20 requests per minute. Free-key rate limits are 100 requests per minute and 100,000 tokens per minute for the Embedding and Reranker APIs. Jina AI has been part of Elastic since October 2025.

Voyage AI, owned by MongoDB since 2025, gives every account free tokens on its current models: 200M tokens each on voyage-4-large, voyage-4, voyage-4-lite, voyage-context-4 and voyage-code-4, 200M tokens on the rerank-3 and rerank-3-lite rerankers, and 200M text tokens plus 150B pixels on the multimodal models. voyage-finance-2, voyage-law-2 and voyage-code-2 get 50M. Older models and Batch API usage get no free tokens.

Cohere creates a free trial API key automatically when you sign up. Trial keys can use all of Cohere's models and APIs, but they are capped at 1,000 API calls per month and have per-endpoint rate limits, such as 20 requests per minute for Chat and 10 for Rerank. Cohere says trial keys may not be used for production or commercial purposes; for that you complete the Go to Production form and pay per use.

ElevenLabs' Free plan includes 10,000 credits per month, which ElevenLabs puts at about 10 minutes of text-to-speech, plus 3 custom voice slots for Voice Design. Free-plan output has no commercial license: you may publish it only non-commercially and must credit ElevenLabs with "elevenlabs.io" or "11.ai" in the title. Commercial rights and voice cloning start with the Starter plan at $6 a month.

Pollinations provides an OpenAI-compatible API at gen.pollinations.ai for text, image, video, audio, realtime voice and embeddings. Generation requests need an API key from enter.pollinations.ai and are billed in Pollen, which you buy (one dollar buys one Pollen) or earn for free as 'Quest Pollen' by completing quests such as creating your first key (0.25 Pollen) or trying each model type. Quest Pollen only works on models not marked paid-only. Legacy raw publishable keys are limited to 1 Pollen per IP per hour, and the model catalogue endpoints are public without auth.

Cloudflare Workers AI includes a free allocation of 10,000 Neurons per day on both the Workers Free and Workers Paid plans, resetting at 00:00 UTC. On Workers Free, requests fail once the daily cap is reached; on Workers Paid, usage beyond the allocation costs $0.011 per 1,000 Neurons. A handful of newer frontier models (for example Kimi, GLM-5.x and DeepSeek V4) are not available on the free plan at all.

NVIDIA's official FAQ says free NVIDIA Developer Program members get free access to NIM API endpoints for prototyping at build.nvidia.com, plus downloadable NIM microservices for research, development and testing on up to 16 GPUs. Production use requires an NVIDIA AI Enterprise license (90-day free trial, then from $4,500 per GPU per year). NVIDIA doesn't publish fixed credit amounts or rate limits for the hosted endpoints; staff describe limits as varying by model, use case and overall traffic.

Cerebras no longer offers a permanently free tier; new accounts get a one-time $5 promotional credit after adding a verified payment method. Credits expire 30 days after being granted, can be used across all public models, and you aren't charged unless you buy more. Free Trial usage is rate-limited per model (5 requests per minute and 1M tokens per day on gpt-oss-120b), and your first purchase moves you to the pay-as-you-go Developer tier.

Per Hugging Face's current pricing docs, Free accounts receive no monthly Inference Providers credits and must purchase credits to use the service. PRO users ($9/month) and Team or Enterprise organizations ($2 per seat) get $2.00 of compute credits each month, usable across Hugging Face compute services. Hugging Face passes provider prices through with no markup.

DeepSeek's official pricing page lists two API models, deepseek-flash (now served by DeepSeek-V4.1-Flash-1017) and deepseek-v4-pro, with pay-as-you-go token pricing that is halved during off-peak hours. The docs mention a 'granted balance' that is spent before your topped-up balance, but publish no free-credit amount or eligibility rules for new users. The only free usage DeepSeek documents is a temporary promotion on its new Search and Fetch tools, which become billable on 1 November 2026 (Beijing time).

xAI's current API docs describe prepaid credits, auto top-up and promo-code redemption but do not mention a signup credit or a $150/month data-sharing credit program. The only officially documented free API credit was $25 per month for everyone during the public beta announced on 4 November 2024, which ran until the end of 2024. xAI's enterprise FAQ says it may, in some instances, offer free credits in exchange for permission to train on business data, but publishes no amount or sign-up route. Third-party sites report the later $25 signup and $150/month data-sharing credits ended; we could not confirm this on an official page.

Fireworks AI's pricing page and docs do not mention free signup credits; Fireworks runs on prepaid credits, so new users add a payment method and buy credits, with usage deducted from the balance. The only free offer listed officially is free LoRA SFT fine-tuning of Muse Glimmer 30B via the UI and Serverless Training API until 11/30 (jobs started before December 1, 2026, Pacific time), with no credit card required. Accounts with no payment method or no credits are limited to 10 requests per minute.