Free AI APIs

Which AI APIs you can still call for free, what each one asks for (a card, a phone number, nothing), and which model IDs actually answered when we called them.

14 providers tracked · latest verification (1 of 14 checked that date) · 33 articles in this category

2 Free after a phone check Zhipu GLM, Gonka Broker
1 Trial key only Cohere
2 Card required Cerebras, SambaNova Cloud

Which one should you use?

Fastest replies (chat, voice, agents) Groq Free, works The fastest generation we have measured on any free tier. Checked 2026-10-11 · review due Long documents, PDFs, images Google Gemini Free, works Flash models are free with a very large context window; expect slow first tokens on thinking models. Checked 2026-10-09 · review due Trying many models with one key OpenRouter Free, works One OpenAI-compatible key for every :free model; the free list changes often, so check the model table. Checked 2026-10-09 · review due Code completion and small multilingual models Mistral Free, works Codestral and the Ministral models answer on a free key; the small and medium models do not. Checked 2026-10-09 · review due Large open models (Nemotron, gpt-oss) for free NVIDIA NIM Free, works Hosts big models free for prototyping, but most of its listed models no longer answer. Checked 2026-10-11 · review due Chinese-language apps Zhipu GLM Free, phone check Free Flash models with strong Chinese; sign-up needs a phone number. Checked 2026-10-09 · review due DeepSeek V4 Flash without paying DeepSeek Gonka Broker Free, phone check Free monthly tokens on DeepSeek, GLM and MiniMax models, after a phone check. Checked 2026-10-09 · review due AI inside a Cloudflare Worker Cloudflare Workers AI Free, works Free daily Neurons on the edge; useful if your app already runs on Workers. Checked 2026-10-11 · review due Embeddings and reranking for RAG Cohere Trial only Trial key includes Embed and Rerank; not for production traffic. Checked 2026-10-09 · review due

At a glance

The short version. Quotas, rate limits and every model ID are in the full comparison.

ProviderStatusCard?Phone?Models that answeredFastest first tokenChecked
Groq Free, works No — 4 0.88 s qwen/qwen3.8-27b review due
Google Gemini Free, works No — 10 1.25 s gemini-3.5-flash-lite review due
Mistral Free, works No — 12 0.69 s ministral-8b-latest review due
OpenRouter Free, works No — 8 1.22 s nvidia/nemotron-3-super-120b-a12b:free review due
NVIDIA NIM Free, works No — 13 0.54 s nvidia/nemotron-3-super-120b-a12b review due
Zhipu GLM Free, phone check No Yes 3 0.65 s glm-4-flash-250414 review due
Gonka Broker Free, phone check No Yes 3 2 s deepseek-ai/DeepSeek-V4-Flash-0731 review due
Cloudflare Workers AI Free, works — — 6 — review due
Cohere Trial only — — — — review due
Cerebras Card required Yes — — — review due
SambaNova Cloud Card required Yes — — — review due
DeepSeek API Closed since 2026-08-31 Yes — — — review due
Together AI Closed since 2026-08-31 Yes — — —
GitHub Models Closed since 2026-07-30 — — — — review due

"—" means not stated by the provider or not measured by us.

Tests, one provider at a time

Roundups and head-to-head tests

Running agents on free tiers

One chat completion is the wrong unit for an agent. These tests count rounds, tokens and the errors that only show up in a loop.

Speech, vision and safety APIs

Search, scraping and gateways

More in this category

How this page stays current

Status, card and phone rules, model counts and first-token times come from one data file per provider. We test selected APIs with a free key and check other claims against provider documentation. Each row shows its own verification date; that date is not the last time we edited this page.