Join our Newsletter — 33% off our NHI Course
Home› Guides› LLM API Free Tiers and Credits Compared: Free…
Guide AI Platforms

LLM API Free Tiers and Credits Compared: Free Access, Credits and Conditions by Provider

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 15 November 2026 13 min read
On this page

Free access to LLM APIs comes in several forms, such as free tiers, trial keys, discounted modes that pay with your data, and credits bundled with chat subscriptions, and each carries its own limits and data terms. This page compares what eight API providers document about free and credited API use: which models can be called at no charge, what the free tier costs you in data use or rate limits, which subscriptions include monthly API credits, and what happens when the free allowance runs out. It is written for developers prototyping on a budget, teams deciding which keys staff may create, and buyers checking whether "free" use is acceptable for company data. Providers are listed alphabetically and order is never a ranking. Where a provider's pages state nothing for a field, the page says so.

Last checked: 2026-11-15.

This is an independent reference page from NHI Mgmt Group based on public provider documentation; it is not sponsored or endorsed by any provider. Free tiers and credit offers change often, so confirm the current terms on each provider's pages, linked in the sources list.

Key takeaways

  • Free API access takes four forms. Google lists free-of-charge models, Cohere issues free evaluation keys, Mistral AI offers a free Experiment plan, and Meta offers a discounted Contributor mode; Anthropic and OpenAI instead attach monthly API credits to some paid chat plans.
  • Several free routes pay with your data. Google states free tier content is used to improve its products, Mistral AI states Experiment plan requests may be used to train its models, and Meta's Contributor mode grants permission to train on prompts and responses.
  • Subscription credits expire every cycle. Anthropic (Max, Ultra, Team) and OpenAI (Pro, Business) both state unused monthly API credits expire at the end of the billing cycle and are spent before purchased credits.
  • Cloud trial credits do not always cover AI. Google Cloud's $300 Free Trial credit cannot pay for Gemini API costs in AI Studio, according to Google's trial page.
  • Free use is rate limited harder. Cohere evaluation keys allow 1,000 API calls a month, Mistral AI describes its free tier limits as restrictive, and Meta's Contributor variants run at 150 requests a minute against 3,000 for standard mode.

What counts as a free tier or credit on this page?

This page covers the provider's own developer API, not its consumer chat app. A free tier is API access that costs nothing at the point of use, usually with tighter limits. A trial or evaluation key is a free key restricted in volume. A data-for-discount mode lowers or removes the price in exchange for permission to use your content for training. Credits are a money balance granted by the provider, for example with a subscription, a promotion or a cloud trial. Free chat app plans are compared separately in AI Chatbot Free Plans Compared.

Free API access at a glance

As of 2026-11-15, from each provider's documentation.

ProviderFree or credited routeCost to youMain conditionSource
AnthropicMonthly API credits on Max, Ultra and Team plansIncluded with the paid planPlan active for seven days; Free, Pro and Enterprise not eligibleMonthly API credits
CohereEvaluation (trial) keysFree1,000 API calls a month and lower per-minute limitsRate Limits
DeepSeekGranted balance (amount not stated)Not stated by DeepSeekGranted balance is used before topped-up balanceModels and Pricing
Google (Gemini API)Free tier on listed modelsFreeContent used to improve Google products; free tier rate limitsGemini API pricing
Meta (Model API)Contributor mode on Muse Spark 1.2 and 1.3Discounted, not free: $0.10 input, $0.20 output, $0.002 cached input per 1M tokensMeta may train future models on your prompts, code, files, context and responsesPricing and rate limits
Mistral AIExperiment planFree, no credit cardVerified phone number; requests may be used to train Mistral's modelsExperiment plan
OpenAIMonthly API credits on Pro and Business plans; Free usage tierCredits included with the paid planSubscription paid, active and at least seven days old; Free tier needs an allowed geographyAPI credits for ChatGPT subscribers
xAIPromo codes on the spend management pageNot stated by xAIPromo code requiredManage billing

OpenAI's rate limit guide lists a Free usage tier for users in an allowed geography with a $100 monthly usage limit, while its pricing page lists no free token allowance. Read the Free tier as a limit level, not as granted tokens.

Which models can you call for free?

  • Anthropic: subscription credits cover "any available Claude model on the Claude Platform", through the Messages and Message Batches APIs, the Console Playground, Claude Managed Agents and the Claude Agent SDK.
  • Cohere: trial keys can call the Chat models listed on the rate limit page (Command A, Command A+, Command A Reasoning, Translate and Vision, Command R+, R and R7B, North Mini Code) at 20 requests a minute, plus Embed, Rerank and other endpoints. Cohere asks customers to contact sales before using the newer models in production.
  • DeepSeek: the granted balance pays for usage at normal prices on deepseek-flash and deepseek-v4-pro; no model restriction is stated.
  • Google: free-of-charge models on the Standard tier include Gemini 3.8 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini Embedding 2 and Gemma 4, among others. Gemini 4 Argon, Gemini 3.1 Pro Preview and the image and Lyria models are paid only. Some models are free on Standard but not on Batch.
  • Meta: only muse-spark-1.3-contributor and muse-spark-1.2-contributor offer Contributor mode; Muse Wonder, Muse Ember and Muse Voice Transcribe do not.
  • Mistral AI: the Experiment plan page does not list models; frontier model access is listed as a Scale plan feature.
  • OpenAI: subscription credits cover paid API Platform usage including the Responses, Agents and Decisions APIs, GPT-Live, Realtime and Images APIs, and built-in tools such as web search, file search and Code Interpreter. Codex usage through API keys is excluded.
  • xAI: no free model access is stated on the pricing page. Grok 4.7 Fast is not on the public API; the pricing page notes that Grok Build's free tier does not include it.

API credits bundled with subscriptions

As of 2026-11-15. Two providers document monthly API credits attached to chat subscriptions. The other six providers' pages checked describe no equivalent.

ItemAnthropicOpenAI
Eligible plans and amountsMax $100 a month; Ultra $200; Team $20 per Standard seat and $100 per Premium seat, pooled, capped at $500Pro $100, $200 or $500 by Pro tier; Business $20 per Standard seat and $100 per Premium seat, pooled, capped at $500
Not eligibleFree, Pro, EnterpriseFree trials, complimentary subscriptions, Enterprise; downgrading to Plus, Go or Free removes the link
Waiting periodSeven days on an eligible planSubscription at least seven days old
ExpiryUnused credits expire at the end of each billing cycleUnused credits expire at the end of the billing cycle and do not roll over
Order of useBefore purchased creditsBefore purchased credits
ExcludedInteractive Claude Code, extra usage in Claude apps, Claude on Amazon Bedrock, Google Cloud Vertex AI or Microsoft FoundryCodex usage through API keys; credits do not raise ChatGPT or Codex limits
LinkingOne Console organisation per plan; change via supportOne API organisation; same email as the subscription; change via support
When exhaustedRequests stop until next credits unless purchased credits or auto-reload existHTTP 429 credit_balance_exhausted unless paid credits or Auto Recharge exist
SourceMonthly API creditsAPI credits for ChatGPT subscribers

Cloud trial credits

Google Cloud's Free Trial gives new customers "a $300 Welcome credit to spend over 90 days", for those who have never been paying users of Google Cloud, Google Maps Platform or Firebase. The trial page states the credit cannot pay for Gemini API in AI Studio costs or managed generative AI partner models, and that it cannot pay for Gemini platform services for trials or upgrades on or after 16 November 2026. When the credit is spent or 90 days pass, the trial billing account closes, followed by a 30-day grace period before trial resources are deleted. Other cloud trial programmes were not checked for this page.

Is my data used for training on free access?

As of 2026-11-15. Paid terms are covered in depth on LLM Data Retention and Training-Use Policies by Provider.

ProviderFree or credited routeTraining or product improvement statementSource
AnthropicSubscription API creditsNot stated on the credits page; credits are consumed on the Claude Platform under its API termsMonthly API credits
CohereEvaluation keysNot stated on the rate limit pageRate Limits
DeepSeekGranted balanceNot stated on the pricing pageModels and Pricing
Google (Gemini API)Free tierFree tier: content used to improve Google products; paid tier: not usedGemini API pricing
Meta (Model API)Contributor modeYou grant Meta permission to use prompts, code, files, context and responses to train future Meta models; standard mode content is not usedPricing and rate limits
Mistral AIExperiment plan"API requests made under the Experiment plan may be used to train Mistral's models"Experiment plan
OpenAISubscription API creditsNot stated on the credits page; the benefit is subject to API terms and policiesAPI credits for ChatGPT subscribers
xAIPromo codesNot stated on the billing pageManage billing

What limits apply to free use?

  • Anthropic: no separate free tier limits are stated; new or low-history organisations may start in the Evaluation tier with lower limits.
  • Cohere: 1,000 calls a month on evaluation keys; 20 Chat requests a minute; Rerank 10 a minute against 1,000 on production keys.
  • DeepSeek: the same concurrency limits apply to all accounts: 2,500 for deepseek-flash and 500 for deepseek-v4-pro.
  • Google: the Free tier is the lowest rate limit tier; per-model values are shown in AI Studio. Grounding with Google Search is free up to 500 requests a day on Gemini 2.5 Flash and Flash-Lite (shared), and not available on the free tier for Gemini 2.5 Pro and Gemini 3 and newer.
  • Meta: Contributor variants are limited to 150 requests and 3,000,000 tokens a minute.
  • Mistral AI: the free tier has "restrictive rate limits" meant for testing; values are shown at console.mistral.ai/limits.
  • OpenAI: the Free usage tier carries a $100 monthly usage limit; subscription credits remain subject to the organisation's rate limits.
  • xAI: every team starts at Tier 0, which still lists 50M tokens a minute for grok-4.7.

Full tier tables are on LLM API Rate Limits and Usage Tiers Compared.

What happens when free credit runs out?

As of 2026-11-15.

ProviderBilling model after free useBehaviour at zero balanceSource
AnthropicPrepaid credits (expire one year from purchase, non-refundable), optional auto-reload; invoicing through salesAPI calls and Playground stop until credits are addedHow do I pay for my Claude API usage
CohereProduction keys, paidNot stated on the rate limit pageRate Limits
DeepSeekTopped-up balance; off-peak prices are half the peak priceNot stated on the pricing pageModels and Pricing
Google (Gemini API)Link a billing account for Tier 1 and aboveExceeding any free tier limit returns an error; spend limits return 429 RESOURCE_EXHAUSTEDRate limits
Meta (Model API)Standard mode per-token pricesNot stated on the page checkedPricing and rate limits
Mistral AIScale plan, pay per use, with higher limitsNot stated on the page checkedExperiment plan
OpenAICredit purchases, Auto Recharge429 credit_balance_exhaustedAPI credits for ChatGPT subscribers
xAIPrepaid credits, auto top-up (minimum $5), invoiced billing on request ($0 invoiced limit by default)Requests rejected once prepaid credits are depletedManage billing

How to choose a free route: factual criteria

  1. Data sensitivity: if prompts may contain personal, customer or confidential data, the routes where providers state training or product-improvement use (Google free tier, Meta Contributor mode, Mistral AI Experiment) need an explicit decision first.
  2. Model needed: check the model is in scope: Google's free list excludes Gemini 4 Argon, and Meta's Contributor mode excludes Muse Wonder and Muse Ember.
  3. Volume: 1,000 calls a month (Cohere) suits evaluation, not a pilot.
  4. Who owns the account: subscription credits from Anthropic and OpenAI belong to the subscriber and link one organisation; personal credits linked to a work organisation blur ownership.
  5. Exit path: know which billing model follows and whether keys, projects and limits carry over to the paid tier.

Where identity and security fit

Free keys are still credentials. They are created quickly, often by individuals outside procurement, and are easy to leave in code or notebooks; a leaked key on an organisation with auto-reload or invoiced billing can spend real money. Track every key in your API Key Management Guide process, see the LLMjacking Guide for abuse patterns, and use provider key guides such as the Anthropic API Keys Security Guide, Google API Keys Security Guide and OpenAI API Keys Security Guide.

Common failure patterns

  • Prototyping with real data on a training-use tier: the free route is chosen for cost and nobody reads the data statement.
  • Assuming cloud trial credit covers the AI bill: Google's trial page excludes Gemini API in AI Studio costs.
  • Planning around credits that expire: Anthropic and OpenAI subscription credits do not roll over, and Anthropic prepaid credits expire one year after purchase.
  • Personal credits on a work organisation: when the subscriber leaves or downgrades, new grants stop and the workload loses its budget.
  • Treating a credit 429 as a rate limit: OpenAI's credit_balance_exhausted will not clear with backoff.

Practitioner checklist

  • List every free, trial and credited API account in use, with its owner and linked organisation.
  • Record each route's training or product-improvement statement and approve it against your data classification.
  • Keep production traffic off evaluation keys and free tiers.
  • Note credit expiry dates and the order in which granted and purchased credits are spent.
  • Set spend limits or top-up caps before enabling auto-reload, Auto Recharge or invoiced billing.
  • Alert on credit and quota errors separately from rate limit errors.
  • Rotate or revoke free keys when the evaluation ends.

Standards and references

Related NHI Mgmt Group resources: LLM API Rate Limits and Usage Tiers Compared, LLM Pricing Tracker, AI Chatbot Free Plans Compared, LLM Data Retention and Training-Use Policies by Provider.

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 15 November 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.