How Tokens Work
Everything about Free.ai tokens — how they work, what they cost, and how to get more.
توکي څه دي؟
توکي د Free.ai په اړه نړیواله پیسه ده. هر وسیله - خبرې اترې، د انځور نسل، متن-د-خبرو، ژباړې، او نور - د توکو لګښتونه. یو توکي په نږدې توګه د انګلستان د متن 4 کرکټرونه برابروي.
د توکو په اړه فکر وکړئ لکه د آرکیډ په توګه کریډیټ: تاسو د وړیا لپاره ورځنۍ پیسې ترلاسه کوئ، او تاسو کولی شئ نور یې واخلئ که تاسو ورته اړتیا لرئ.
وړيا توکي
| نااشنا (هېڅ ګڼون) | ورځ |
| برېښليک کره کول | د بشپړ 30,000 / ورځ پوټ خلاصوي |
| وړیا حساب (هره ورځ نوي کول) | ورځ |
| د حوالې بونس | 30,000 for you + your friend |
د ورځني توکي په منځنۍ شپه UTC بیا تنظیم او په اتوماتيک ډول پلي کیږي کله چې تاسو تولید کوئ؛ د حساب توازن شمیره یوازې د پیرود شوي توکي ښیې. کله چې ستاسو ورځنی نل کارول کیږي، تاسو کولی شئ په $ 5 کې نور پیل کړئ.
د توکو په واسطه د توکو لګښتونه
| توکي | لګښت | بېلګه |
|---|---|---|
| (خپل کوربه) Chat AI | اوسني توکي | هر استوزه ~100 توکي |
| (GPT-4) Chat د | ~325 توکي/msg | د نمونې سره توپیر لري |
| انځور جوړول | ۱۰۰۰ توکي | د انځور په سر |
| ويډيو جوړول | 30,000 tokens | Per clip (up to 12s; +2,500/s beyond) |
| ليکنه وينا ته | لوښه | 100 chars = 25 توکي |
| وينا ليکنې ته | په سېکېنډ کې ۴ | 1 دقیقې = 240 توکي |
| د سندرې جوړول | 5,000+ توکي | په هر پلنيوي |
| ژباړه | اوسني توکي | ~ 50 توکي په هر فقره |
| ليکنه / کوډ | اوسني توکي | ~ 200 توکي په هر نسل |
د ځان کوربه vs بهرنی
ځان کوربه بېلګه (Qwen 2.5, Kokoro, Whisper, SDXL) run on our GPU servers. They cost actual tokens used - no markup. These are the cheapest option.
بهرني ماډلونه (GPT-4, Claude, Gemini) are accessed via their providers. We convert their USD pricing to tokens at competitive rates.
ډېرې پوښتنې
Tokens are the universal currency on Free.ai. Every tool -- chat, image generation, text-to-speech, translation, and more -- costs tokens. One token roughly equals 4 characters of English text. Think of tokens like credits at an arcade.
Anonymous users get 6,000 tokens per day. Free account holders get 30,000 tokens per day. Referrals earn 30,000 bonus tokens for both you and your friend (one day's worth of the daily pool).
Daily tokens reset at midnight UTC. Any unused daily tokens do not roll over to the next day. Purchased tokens and bonus tokens have no daily expiration.
AI Chat with self-hosted models costs actual tokens used (about 100 per message). Image generation costs 1,000 tokens per image. Video generation costs 30,000 per clip on most models. A 3D mesh costs 45,500 on TRELLIS, 11,000 on TripoSR. Text-to-speech costs 1 token per 4 characters. See the cost table on this page for all tools.
Self-hosted models run on our GPU servers -- you pay actual tokens used. External models (GPT-4, Claude, Gemini) are accessed via their providers at competitive rates.
We convert the provider's per-token USD cost to our token currency. Each model's exact token cost is shown on its page before you use it.
Yes. Visit the pricing page to purchase token packages or subscribe to a plan. Paid plans include higher daily limits and bulk token purchases at lower per-token rates.
Auto top-up automatically purchases additional tokens when your balance drops below a threshold you configure. This prevents running out of tokens mid-task. Set your threshold and top-up amount in account settings.
Yes, within your daily token limit. Self-hosted models like Qwen 2.5 7B, Kokoro TTS, and faster-whisper STT deduct from your daily free allowance with no markup. They are the cheapest way to use AI on Free.ai.
Yes. Image generation costs a flat 1,000 tokens per image. Video costs 30,000 tokens per clip up to 12 seconds, then 2,500 more per second beyond that. The flat per-clip price is there because a render's GPU cost is dominated by model load and diffusion steps rather than clip length: a self-hosted clip averages about six and a half minutes of GPU time. Cheaper models are priced lower (CogVideoX-2B is 11,500), heavier ones higher (HunyuanVideo is 47,000). On a free account video is capped at one clip every two days, and that clip is your whole 30,000-token daily pool for the day you spend it. Both image and video are fixed regardless of the prompt, unlike chat which varies with message length.
Use self-hosted models instead of external ones, keep prompts concise, lower the max tokens setting when you don't need long responses, and use the smaller 7B models for simple tasks where the 72B model is not necessary.
When your daily free tokens are exhausted, you can either wait for the next day's reset at midnight UTC, purchase additional tokens, or upgrade to a paid plan for higher limits. The tool interface shows your remaining daily balance.