Skip to content
API application guidance, personal key tests and tool setup

Find an LLM API that fits.
Choose your next step.

Compare official free-access requirements, run one browser-direct request with your own key, or review the protocol and keyless setup notes for a tool. These are parallel choices, not a required sequence.

I need a key

Check free-access, account and regional requirements before applying through the official path.

Find an API I can access

I have a key to test

Send one browser-direct request. The result applies to that call and may use quota.

Test my own key

I want to set up a tool

Review protocol compatibility, then copy a keyless config or setup notes. Testing first is optional.

Review tool setup
6 reviewed providers24 model routes2 latest sampled checks passed0 plaintext keys on server
Describe the outcome

What does your free API need to do?

FreeToken maps common needs to reviewed catalog fields. No external AI call is used.

Choose an example or type a requirement in your own words.

Curated launch set

Useful models, with the access path attached

24 reviewed catalog matchesOpen all filters →

Capabilities and speed classes are catalog hints, not benchmark rankings or client compatibility guarantees.

NVIDIA
Latest observation is staleObserved:

GLM-5.2

z-ai/glm-5.2

Long-horizon coding and reasoning candidate in NVIDIA's catalog.

Why it fits: Developer access · OpenAI compatible

Access
Developer API access
Protocol
OpenAI compatible

Official conditions reviewed
Recheck official terms before applying; this is not a live service check.

Gemini
Latest sampled check found quota exhaustedObserved:

Gemini Flash

gemini-flash-latest

General-purpose multimodal model through Google AI Studio.

Why it fits: No card required · Gemini native

Access
Account-specific free tier
Protocol
Gemini native

Official conditions reviewed
Recheck official terms before applying; this is not a live service check.

Groq
Latest sampled check passedObserved:

Qwen3.8 27B

qwen/qwen3.8-27b

Groq Preview successor for coding and reasoning. Check the latest route observation separately.

Why it fits: No card required · OpenAI compatible

Access
Preview model on the Groq Free plan
Protocol
OpenAI compatible

Official conditions reviewed
Recheck official terms before applying; this is not a live service check.

SiliconFlow
No platform check is configured

DeepSeek R1

deepseek-ai/DeepSeek-R1

Reasoning model available through SiliconFlow's compatible API.

Why it fits: No card required · OpenAI compatible

Access
Credits or free-model policy
Protocol
OpenAI compatible

Official conditions reviewed
Recheck official terms before applying; this is not a live service check.

Optional Watchlist

Follow meaningful changes, not every check.

Choose a provider or exact model, then receive recovery, sustained failure and access-rule updates on your schedule.

Build a Watchlist →
Optional encrypted Vault

Keep a compatible key workflow when you choose.

Import supported local formats or encrypted backups, then export keyless profiles where the tool guidance supports them.

Open Import & export →
Personal key test

Run one direct request with your own credentials.

Select a model and test it browser-direct. The result applies to that request and can differ by account, region and remaining quota.

Try a model →
Cross-device by design

Your setup should follow you, not stay trapped in one browser.

Browse anonymously. Sign in only if you want the optional client-encrypted Vault, saved configurations, history or alerts. It is never a gate to discovery, key testing or tool guidance.

Optional continuity

Use the Vault only when it helps your workflow.

Discovery, official application guidance and a browser-direct personal test do not require a stored key. Tool compatibility is documented per route rather than assumed for every client.

Review Vault boundaries →