This third-pass expansion adds 161 new candidate offers from 111 independent source hosts, on top of the earlier 59-row report — spanning LLM and gateway APIs (57), compute, sandbox and agent hosting (27), image/video/avatar (22), search/scraping/browser (21), voice (17), document/OCR (10) and selective credit programs (7), as of 27 September 2026. Every row is a lead to verify on signup, not an account-tested offer; tokens, requests, dollars, minutes and GPU credits are deliberately never ranked on one scale.
52 of the 161 leads explicitly state that no credit card or payment method is needed — e.g. Daytona's $200 CPU-sandbox credit (GPU sandboxes excluded) — buildaicurrent.com
The biggest recurring token allowance found is LongCat AI's public beta: 50M Flash-Lite plus 500K Chat/Thinking tokens per day, beta-only and revocable — newreleases.io
Only 49 leads have two or more supporting pages; 105 rest on a single launch page, commit or community post and are badged “verify on signup” — e.g. nCompass's $100 rate-limit-free inference credit sits on one YC launch page — ycombinator.com
The 7 selective programs (AWS AI safety compute up to $100K, Google Cloud startup credits, OpenAI academic access and more) are application-gated grants, never per-user entitlements — milik.ai
| Provider | Category | Reported free offer | Key conditions | Type | Evidence | Date | Source |
|---|
Discovery catalogue of 161 candidate free offers gathered 2026-09-27 from ~8,000 web search results, filtered to 111 independent source hosts; one row per provider within category. Each row reports the free allowance as stated by its source page (tokens, requests, dollars, minutes or credits — units are not comparable and are never summed). Offer type and no-card flags are inferred from source wording; every row is a lead to verify on signup, not account-tested. Long condition texts are truncated for space; follow the source link for full terms. Additions beyond the previous 59-row report; dates shown as unverified where the source gave none.