Unofficial · not affiliated with any providerFree AI API credits, tracked weekly · Last updated: Oct 9, 2026
Last updated: · 17 active offers · 2 recently ended
Free AI API Credits & Free Model Offers (Updated Weekly)
Free tiers, trial credits and limited-time free models from 17 providers, including Chinese vendors such as Qwen, Zhipu GLM, StepFun and LongCat. Each entry lists what's free, the limits, how to claim it the official way, the deadline and whether you need a card, with a link to the official source.
Unofficial, not affiliated with any provider; offers change often — check the official link.
Free use of Qwen 3.8 Max, Qwen 3.8 Flash and Wan3.0-video through GMI Cloud's OpenAI-compatible API during the promotion. Enterprise teams can also apply for voucher credits using a form on the same page.
Limits
Your account balance must be at least $10. The $10 isn't spent and stays in your balance. GMI raised rate limits on Oct 7, 2026 but hasn't published the exact numbers.
How to claim
Sign up at GMI Cloud → Make sure your balance is at least $10 (top up if you're new) → Call the models through the OpenAI-compatible API with your GMI key
Deadline / expiry
Oct 11, 2026 (per GMI's reply on X; timezone not stated)
Extended?
Yes. Free access was extended by 7 more days on Oct 7, 2026, with higher rate limits.
Commercial use?
Not specified. The offer is "for legitimate use only" and subject to GMI Cloud and Alibaba Cloud Terms of Service and Acceptable Use Policy.
During the promotion, calls to Qwen3.8-Flash in Qoder International products cost 0 Credits (the Credit rate drops from 0.1x to 0.0x). Free accounts and accounts with zero Credits qualify.
Limits
Only works inside Qoder products (Qoder IDE, JetBrains plugin, CLI, QoderWake, Cloud Agents, Mobile, web app); it isn't an exportable API. Individual accounts only; Teams and Enterprise are excluded. Qoder warns responses may be slower at peak times. Usage caps aren't published.
How to claim
Create or sign in to a Qoder International individual account → Select Qwen3.8-Flash as the model → Use it normally; nothing to claim
Deadline / expiry
Extended past Sep 30, 2026; Qoder says it will post the end date on the page in advance
Extended?
Yes. Originally ended Sep 30, 2026 23:59:59 (UTC+8).
Commercial use?
Not specified on the official page; check the provider's terms.
StepFun announced a week of free access to Step 5 Preview starting Oct 8, 2026 in coding tools including OpenCode, Cline, Nous Research (Nous Portal) and Kilo Code. Step 5 Preview is also listed on OpenRouter, but the announcement doesn't say OpenRouter access is free.
Limits
Free use happens inside the partner tools. Per-tool usage caps aren't published.
How to claim
Open one of the partner tools (OpenCode, Cline, Kilo Code or Nous Portal) → Select Step 5 Preview in the model picker → Use it during the free week
Deadline / expiry
About Oct 15, 2026 ("a week" from Oct 8; exact end time not published)
Extended?
No (as of this check)
Commercial use?
Not specified on the official page; check the provider's terms.
Ongoing free tiers with rate limits. No end date has been announced.
OngoingAlways free Last verified:
Cloudflare Workers AI: Llama, Gemma, Qwen, GLM-4.7-Flash, gpt-oss, FLUX image models, Whisper and more (per-model…
What's free
Every account gets 10,000 Neurons per day at no charge, on both the Workers Free and Workers Paid plans. Neurons measure GPU compute; each model has a published Neuron price per million tokens, per image tile/step or per audio minute.
Limits
10,000 Neurons/day. Limits reset daily at 00:00 UTC; when you exceed them on Workers Free, requests fail with an error. Above the free allocation, Workers Paid costs $0.011 per 1,000 Neurons. Some models need a paid billing method, including @cf/moonshotai/kimi-k2.6, kimi-k2.7-code, @cf/zai-org/glm-5.2, glm-5.3, glm-5.3-flash, deepseek-v4-flash-0731 and deepseek-v4-pro-0813.
How to claim
Create a Cloudflare account → Open Workers AI in the dashboard and create an API token, or use the AI binding in a Worker → Pick a model that doesn't require a paid billing method → Watch Neuron usage in the Workers AI dashboard
Deadline / expiry
Ongoing (no end date published)
Extended?
Not applicable
Commercial use?
Not specified on the official page; check the provider's terms.
Card required?
No for the free allocation on Workers Free. Workers Paid (for usage above 10,000 Neurons) is a paid plan.
Google (Gemini API / AI Studio): Gemini 3.8 Flash, 3.7 Flash, 3.6 Flash, 3.5 Flash, 3.5 Flash-Lite, 3.1 Flash-Lite, 2.5 Pro…
What's free
The Gemini API Free tier charges nothing for input and output tokens on the models marked "Free of charge" on Google's pricing page, including Gemini 3.8 Flash, 3.7 Flash, 3.6 Flash, 3.5 Flash, 3.5 Flash-Lite, 3.1 Flash-Lite, 2.5 Pro, 2.5 Flash and 2.5 Flash-Lite. Gemini 3.1 Pro Preview, the Nano Banana image models, Veo, Lyria and Omni Flash are not available on the Free tier.
Limits
Per-model rate limits (requests per minute, tokens per minute, requests per day) are not listed as fixed numbers in the docs. Google says to check your active limits in AI Studio, and daily quotas reset at midnight Pacific time. Batch, Flex and Grounding with Google Search are not available on the Free tier.
How to claim
Open Google AI Studio (aistudio.google.com) and sign in with your Google account → Create an API key in AI Studio → Call a free-tier model such as gemini-3.8-flash → Don't link a billing account if you want to stay on the Free tier (linking billing moves the project to Tier 1)
Deadline / expiry
Ongoing (no end date published)
Extended?
Not applicable
Commercial use?
Google's Gemini API terms call free usage "Unpaid Services". Google uses your prompts and responses to improve its products, and human reviewers may read them, so don't send sensitive or personal data. Read the Gemini API Additional Terms before shipping a product.
Card required?
No. The Free tier only needs an active project; billing is needed for Tier 1 and above.
Groq: Models on GroqCloud (e.g. openai/gpt-oss-120b, gpt-oss-20b, qwen/qwen3.8-27b, Whisper)
What's free
GroqCloud has a Free tier: you can create an API key and call models without adding a payment method. You add a payment method only to upgrade to the Developer tier.
Limits
Groq doesn't publish separate free-tier numbers. Its rate-limits page lists base limits for the Developer plan (for example 30 RPM, 1K requests/day, 8K tokens/minute and 200K tokens/day for openai/gpt-oss-120b) and says your exact limits are on the Limits page of your account. Limits apply per organization.
How to claim
Sign up at console.groq.com → Create an API key → Check your organization's exact limits on the Limits page in settings → Upgrade to Developer only if you need higher limits (a payment method is required)
Deadline / expiry
Ongoing (no end date published)
Extended?
Not applicable
Commercial use?
Not specified on the official page; check the provider's terms.
Card required?
No for the Free tier. Upgrading to Developer requires a credit card, US bank account or SEPA debit.
Mistral AI: Mistral models on the API (e.g. mistral-small-latest)
What's free
New Mistral accounts start in Free mode, with API access on by default and no credit card required. Free mode includes a monthly usage allowance with rate limits.
Limits
Mistral doesn't publish the Free mode amounts in its docs. Your included monthly usage and rate limits are shown on the Limits page in Studio. Pay-as-you-go extends usage beyond the included amount.
How to claim
Create a Mistral account and open Studio (console.mistral.ai) → Go to API Keys and click Create new key → Copy the key (it's shown only once) → Test with a model such as mistral-small-latest
Deadline / expiry
Ongoing (no end date published)
Extended?
Not applicable
Commercial use?
Not specified on the official page; check the provider's terms.
NVIDIA (NIM / build.nvidia.com): Models in the NVIDIA API catalog at build.nvidia.com
What's free
Members of the free NVIDIA Developer Program get free access to NIM API endpoints for prototyping, and can download NIM microservices for research, development and experimentation on up to 16 GPUs.
Limits
Rate limits for the hosted endpoints aren't published in NVIDIA's NIM FAQ. Access lasts as long as your program membership.
How to claim
Join the free NVIDIA Developer Program → Open build.nvidia.com and pick a model → Generate an API key from the model page and call the hosted endpoint → For production, get an NVIDIA AI Enterprise license (a free 90-day trial is offered)
Deadline / expiry
Ongoing while you're a program member
Extended?
Not applicable
Commercial use?
No. Developer Program access is for prototyping, research, development and testing only. Serving real end users counts as production and needs an NVIDIA AI Enterprise license.
OpenRouter: Models whose ID ends in :free (the list changes often)
What's free
OpenRouter offers free variants of some models; their IDs end in ":free". Requests to these variants cost nothing, but platform rate limits apply.
Limits
20 requests per minute. 50 requests per day if you've bought less than 10 credits in total; 1,000 requests per day once you've bought at least 10 credits (all-time). Daily counters run on UTC days. If your account balance is negative, even free models can return 402 errors. The GET /api/v1/key endpoint shows your free-model daily counter.
How to claim
Create an OpenRouter account and an API key → Browse free models on the official Free Models collection page → Use the model ID with the :free suffix → Optional: buying 10 credits raises the daily free-model limit from 50 to 1,000 requests
Deadline / expiry
Ongoing (individual free models come and go)
Extended?
Not applicable
Commercial use?
Not specified on the official page; check the provider's terms.
Card required?
No for the 50-requests/day level. The 1,000/day level requires buying at least 10 credits.
Inception's Mercury Decide has a free endpoint on OpenRouter, next to a new paid zero-data-retention endpoint ($0.02/M input; output and cached input free).
Zhipu AI / Z.ai (GLM): GLM-4.7-Flash, GLM-4.5-Flash, GLM-4.6V-Flash (Z.ai); also GLM-4-Flash-250414, GLM-Z1-Flash…
What's free
Zhipu's international platform Z.ai lists GLM-4.7-Flash, GLM-4.5-Flash and GLM-4.6V-Flash as Free for input, cached input and output. The China platform (BigModel, bigmodel.cn) also lists several Flash models as 免费 (free), including the CogView-3-Flash image model and the CogVideoX-Flash video model.
Limits
Each model has a concurrency limit (simultaneous requests) that depends on your account level. You can see your limits in the console; the docs don't publish the numbers. Note that the "限时免费" (limited-time free) labels on BigModel's price table refer to cache storage on paid models, not to the models themselves.
How to claim
Create an account on Z.ai (international) or BigModel (China) → Create an API key in the console → Call a model listed as Free, such as glm-4.7-flash → Check your concurrency limit on the rate-limits page in the console
Deadline / expiry
Ongoing (no end date published)
Extended?
Not applicable
Commercial use?
Not specified on the official page; check the provider's terms.
Alibaba Cloud Model Studio (Qwen): Qwen and other Model Studio models in the Singapore region (international deployment scope…
What's free
When you first activate Model Studio in the Singapore region, Alibaba Cloud gives you a separate free token quota for each eligible model (typically 1,000,000 tokens per model, input and output combined). Dated snapshot versions get their own quota.
Limits
The quota is valid for 90 days from activation, the model's release or your request approval, whichever is later. It covers real-time inference only (not batch, fine-tuning or deployment). Quotas are per model and aren't transferable. An Alibaba Cloud account and its RAM users share one quota, and re-registering doesn't grant a new one. Token Plan and Coding Plan keys don't use the free quota.
How to claim
Complete your Alibaba Cloud account information → Open Model Studio (Singapore) and accept the service agreement; the quota is granted automatically, usually within two hours → Check remaining quota on the model usage page (Free Quota tab) → Turn on "Free Quota Only" to stop the service instead of billing you when the quota runs out
Deadline / expiry
Quota expires 90 days after activation
Extended?
Validity was changed to 90 days for users activating from Sep 8, 2026 (03:00 UTC); earlier users are unaffected
Commercial use?
Not specified on the official page; check the provider's terms.
Card required?
The free-quota page doesn't say; you must complete your account information before activation.
Cerebras: All models in Cerebras Shared Inference (e.g. gpt-oss-120b, qwen-3.8-27b)
What's free
New accounts get $5 in free credits after adding a verified payment method. Cerebras says there's no permanently free tier and no per-model always-free allowance.
Limits
The credits expire 30 days after they're granted. Free Trial rate limits are low; for gpt-oss-120b they are 5 requests/minute, 30K uncached tokens/minute, 1M tokens/hour and 1M tokens/day. Once the credits run out or expire, API access stops until you buy credits.
How to claim
Sign up for Cerebras Cloud → Add a verified payment method (required to activate the API and Playground) → The $5 credit is added; there's no charge until you buy more
Deadline / expiry
Credits expire 30 days after they're granted
Extended?
No
Commercial use?
Not specified on the official page; check the provider's terms.
Not free on their own, but included with a subscription you may already pay for.
ActiveWith paid plan Last verified:
Anthropic (Claude): Claude API, Managed Agents, Agent SDK
What's free
Paid Claude Max and Team plans now include monthly Claude API credits: $100 (Max 5x), $200 (Max 20x), or $20 per Standard seat and $100 per Premium seat on Team (pooled, capped at $500).
Limits
No rollover; credits expire at the end of each billing cycle. They don't cover Claude Code or extra usage in the Claude apps. New subscribers can claim after 7 days. Free, Pro and Enterprise plans aren't eligible.
How to claim
On claude.ai open Settings > Billing (Max) or Organization settings > Billing (Team) → Link a Claude Console organization in the API credits section → Credits appear under Promotional credits in the Console
Deadline / expiry
Ongoing (monthly)
Extended?
Not applicable
Commercial use?
Not specified on the official page; check the provider's terms.
Card required?
No payment method needed on Claude Platform, but you need a paid Max or Team plan.
EndedCline: DeepSeek-V4.1-Flash — Free DeepSeek-V4.1-Flash in the Cline provider. Paused Oct 4, 2026 because of abuse. @cline on X (Oct 4, 2026) ↗
EndedAnt Ling / OpenRouter: Ling-3.0-flash-VL — Two weeks of free access on OpenRouter. Ended Sep 23, 2026, 9:00 AM PT; now paid at discounted prices. @AntLingAGI on X (Sep 23, 2026) ↗
Seen elsewhere, not listed (unverified)
These are circulating on social media, but we couldn't confirm them on an official provider page, so we don't list details or links:
Tencent WorkBuddy (Hy3 / Hy4 preview free quota)
StepFun Step Plan (stackable trial days via invites)
DeepSeek Harness trial credit
Tencent LightVela new-user month
Doubao Work 30-day subscription
Tencent Cloud TokenHub 1M tokens per model (conflicting reports)
Mistral fixed monthly dollar credit
Volcengine Ark (火山方舟) per-model Seedream image quotas (amounts not published in the docs)
SiliconFlow Kolors image model listed as free, and referral coupons (not confirmed on an official page)
ModelScope daily free inference credits
How we pick and verify offers
We only list an offer after checking the provider's own docs, pricing page or official X account. Each entry shows its “Last verified” date.
We don't invent numbers. If a limit or deadline isn't published, we say so.
No gray-hat content: no multi-account tricks, no bulk farming, no key resellers or unofficial relay services.
Frequently asked questions
Are these free AI API offers legitimate?
Every listed offer was checked on the provider's own website, documentation or official X account on the date shown. We link the official source for each one.
How often is this page updated?
Weekly, plus ad-hoc updates when a provider extends or ends an offer. Each entry shows its own 'Last verified' date.
Which free AI API needs no credit card?
According to the official pages: Qoder (Alibaba), Ant Ling via Novita AI, StepFun (阶跃星辰), Nous Research (Nous Portal), Meituan LongCat, Cloudflare Workers AI, Google (Gemini API / AI Studio), Groq, Mistral AI, NVIDIA (NIM / build.nvidia.com), OpenRouter, OpenRouter / Inception, Zhipu AI / Z.ai (GLM). Cerebras requires a verified payment method for its $5 trial.
Can I create several accounts to get more free credits?
No. Providers forbid it, and abuse is a common reason free promotions get paused. We don't list tricks for multiple accounts, bulk farming or reselling keys.