Free tiers in one place, each with its own verification date. LLM API cards use the same provider data as our comparison table. Selected search, app-builder, and hosting cards also read verified provider data; older cards retain their original dates until checked. Each card links to a related article.
Free tiers move. Providers halve allowances, add a card requirement, or stop publishing limits entirely. A figure here is only as good as its date.
Groq
checked · review due- Free API status
- Free, works
- Free allowance
- 1,000 requests/day and 200K tokens/day per model
- Rate limit
- 30 requests/min, 8K tokens/min per model
- Credit card
- No
- Phone check
- Not stated
- Models that answered
- 4
Google Gemini
checked · review due- Free API status
- Free, works
- Free allowance
- Not published — shown only in AI Studio
- Rate limit
- Not published — shown only in AI Studio
- Credit card
- No
- Phone check
- Not stated
- Models that answered
- 10
Mistral
checked · review due- Free API status
- Free, works
- Free allowance
- Not published — shown in your console
- Rate limit
- Not published; small/medium/magistral models get 0 requests/min on free keys
- Credit card
- No
- Phone check
- Not stated
- Models that answered
- 12
OpenRouter
checked · review due- Free API status
- Free, works
- Free allowance
- 50 requests/day on :free models; 1,000/day after $10 of credits ever bought
- Rate limit
- 20 requests/min on :free models
- Credit card
- No
- Phone check
- Not stated
- Models that answered
- 8
NVIDIA NIM
checked · review due- Free API status
- Free, works
- Free allowance
- Free for Developer Program members, for prototyping; no number published
- Rate limit
- Not published in the docs
- Credit card
- No
- Phone check
- Not stated
- Models that answered
- 13
Zhipu GLM
checked · review due- Free API status
- Free after phone check
- Free allowance
- Free Flash models: GLM-4.7-Flash, GLM-4.5-Flash, GLM-4-Flash-250414
- Rate limit
- Concurrency limit per account, shown only in the console
- Credit card
- No
- Phone check
- Yes
- Models that answered
- 3
Gonka Broker
checked · review due- Free API status
- Free after phone check
- Free allowance
- At least 1M tokens/month, renewed on the 1st
- Rate limit
- 6 requests/min per account
- Credit card
- No
- Phone check
- Yes
- Models that answered
- 3
Cloudflare Workers AI
checked · review due- Free API status
- Free, works
- Free allowance
- 10,000 Neurons/day
- Rate limit
- Per-model limits; not compared here
- Credit card
- Not stated
- Phone check
- Not stated
- Models that answered
- 6
Cohere
checked · review due- Free API status
- Trial only
- Free allowance
- 1,000 API calls/month on a trial key
- Rate limit
- 20 chat requests/min (trial key)
- Credit card
- Not stated
- Phone check
- Not stated
- Models that answered
- We hold no key: not called by us
Cerebras
checked · review due- Free API status
- Card required
- Free allowance
- $5 trial credit, expires after 30 days; 1M tokens/day
- Rate limit
- 5 requests/min (trial)
- Credit card
- Yes
- Phone check
- Not stated
- Models that answered
- We hold no key: not called by us. Trial tier lists gpt-oss-120b and qwen-3.8-27b
SambaNova Cloud
checked · review due- Free API status
- Card required
- Free allowance
- Docs list 20 requests/day per model with no payment method; our key got 402 PAYMENT_METHOD_REQUIRED
- Rate limit
- Docs: 20 requests/min per model (free tier)
- Credit card
- Yes
- Phone check
- Not stated
- Models that answered
- Our key now returns 401; not re-tested since 2026-08-31
DeepSeek API
checked · review due- Free API status
- Free API closed
- Free allowance
- No free credit for new accounts on the pricing page
- Rate limit
- Not stated
- Credit card
- Yes
- Phone check
- Not stated
- Models that answered
- Both models return 402 Insufficient Balance
Together AI
checked- Free API status
- Free API closed
- Free allowance
- No free trial; $15 minimum first purchase, $5 minimum later
- Rate limit
- Not stated
- Credit card
- Yes
- Phone check
- Not stated
- Models that answered
- Account read-only until you deposit
GitHub Models
checked · review due- Free API status
- Free API closed
- Free allowance
- Service retired
- Rate limit
- Not stated
- Credit card
- Not stated
- Phone check
- Not stated
- Models that answered
- Retired: playground, catalog and inference API shut down
Render
checked- Free web services
- 750 instance hours/month per workspace
- Sleeps
- After 15 minutes without inbound traffic
- Best for
- Web apps and APIs that can tolerate cold starts
- Cron and workers
- Paid only; no Free instances
Hugging Face Spaces
checked- Free path
- Static Spaces; up to 2 Gradio ZeroGPU Spaces for eligible personal accounts
- Eligibility
- Free personal account in good standing for ZeroGPU
- Other Gradio / Docker
- Paid plan required to create
- Best for
- Public static pages and eligible ZeroGPU demos
ElevenLabs
checked- Free allowance
- 10,000 credits/month
- Free use rights
- Non-commercial use with attribution
- Instant voice cloning
- Starter plan or above ($6/month)
- Best for
- Non-commercial voice prototypes
Coolify
checked- Cost
- Self-hosted Coolify is free; you provide and pay for the server, unless your server has a free allowance
- Resources
- Official install overview: 2 CPU cores, 2 GB RAM, 10 GB free disk; allow more space for apps, images and backups
- What you can deploy
- Apps, databases and one-click services on your own Linux server
- Sleeps / expires
- No Coolify trial clock or imposed sleep; availability depends on your server
- Credit card
- Not needed for the self-hosted software; your server provider may require one
- Commercial use
- Yes; Apache 2.0 open-source license
Cloudflare Workers
checked- Free requests
- 100,000 per day, resetting at midnight UTC
- CPU time
- 10 ms per HTTP request
- Memory
- 128 MB per isolate
- Cron triggers
- 5 per account
- Credit card
- Not required for Workers Free
- Best for
- Edge APIs and short serverless work; not a persistent process
Vercel
checked- Free plan
- Hobby, for personal non-commercial projects
- Included usage
- 4 Active CPU hours, 1M function invocations, 100 GB Fast Data Transfer
- At the limit
- Feature pauses until its usage window resets
- Best for
- Personal Next.js and frontend projects
Exa
checked- Free allowance
- $10 credits every month plus a $10 onboarding bonus
- Resets
- Monthly; onboarding bonus is one-time
- Search rate
- Up to 10 search queries per second on Starter
- Best for
- Semantic search and finding similar pages
Tavily
checked- Free allowance
- 1,000 API credits/month
- Resets
- First day of each month
- After allowance
- Requests stop until reset or upgrade
- Best for
- LLM-ready search and extraction
Brave Search API
checked- Free allowance
- $5 credits each month on a paid plan (about 1,000 Search queries)
- Credit card
- Required, even with $0 prepay
- Standalone free plan
- No
PocketBase
checked- Database
- Embedded SQLite
- Vendor quota
- None for self-hosted software
- Includes
- Auth, files, realtime, admin UI
- Hosting
- Bring your own persistent server
Fly.io
checked 2026-09-14- Free tier
- Nothing standing
- Sleeps / expires
- No free tier for new accounts
- Best for
- Docker, global edge
- The catch
- Free allowances retired in 2024; pay-as-you-go, short trial only
- Free web hosting
- None for new accounts since Oct 2024
- Cold starts
- N/A (no free tier)
Supabase
checked- Database
- 500 MB per project
- Active projects
- 2
- File storage
- 1 GB
- Idle projects
- May pause after 7 days of low activity
Neon
checked- Database
- 1 GB per project
- Projects
- 100
- Compute
- 100 CU-hours/month per project
- Backend
- Managed Auth, 5 GB object storage, Functions
Turso
checked- Databases
- 100
- Storage
- 5 GB across databases
- Rows read
- 500M/month
- Rows written
- 10M/month
Upstash Redis
checked- Data size
- 256 MB
- Commands
- 500K/month
- Bandwidth
- 10 GB/month
- Free databases
- 1 on the pricing comparison
- Inactivity
- FAQ: archive after 30+ idle days; terms permit deletion after 1 idle week
Google Cloud TTS
checked 2026-09-02- Free tier
- Recurring monthly (renews forever)
- Credit card
- Required
- Best for
- High-volume audio on a permanent free quota
- Voice quality
- Very good (WaveNet / Neural2 / Chirp 3)
- Paid rate (cheapest)
- $4/1M chars (Standard and WaveNet), $16/1M (Neural2)
- Free monthly volume
- 4M chars Standard / 1M premium
OpenAI TTS
checked 2026-09-02- Free tier
- No free tier — pay as you go
- Credit card
- Required
- Best for
- Adding voice to an app already calling OpenAI
- Voice quality
- Good, steerable with gpt-4o-mini-tts
- Paid rate (cheapest)
- $15/1M chars ( tts-1 )
- Commercial use on free tier
- Yes (it's all paid)
Alibaba Bailian
checked 2026-08-31- Free tier
- 1M tokens/model
- Credit card
- No
- Speed
- Fast
- Best Model
- Qwen 3.6-Plus
- Paid Pricing
- Very low
- Reasoning
- QwQ (limited)
Dify
checked 2026-08-31- Rate limits
- 200 message credits/day, 5 apps, 5 MB knowledge base
- Paid from
- Free (OSS)
- Best for
- Trying Dify without setup
- Minimal (testing)
- 2 cores
- Recommended (production)
- 4 cores
- High load
- 8+ cores
NotebookLM
checked 2026-09-01- Free tier
- Generous
- Paid from
- Google One AI Premium (~USD 19.99/mo)
- Best for
- Deep doc analysis
- Notebooks
- Unlimited
- Sources per notebook
- 300
- Words per source
- 500,000
Crawl4AI
checked 2026-09-01- Free tier
- Unlimited (your hardware)
- License
- Apache 2.0
- Best for
- High volume, $0 at scale
- Model
- Self-hosted library
- API key
- No
- Deep site crawl
- Yes
Netlify
checked- Free allowance
- 300 shared credits/month with a hard cap
- Metering examples
- 20 credits/GB bandwidth; 15 credits/production deploy
- Database
- Netlify Database is included; usage draws from the credit pool
- At the limit
- Projects pause until the next billing cycle
Northflank
checked 2026-09-09- Free tier
- 2 services, 1 database, 2 cron jobs
- Credit card
- Yes — required for every account
- Sleeps / expires
- No — its pricing page says “always-on-compute”
- Free compute
- 2 services, 2 jobs, 1 addon (database)
- Long / background work?
- Yes — persistent containers and jobs
- Sleeps or expires?
- No sleeping; free plan is framed for hobby use, not production
Cloudflare Pages
checked 2026-09-02- Credit card
- No
- Commercial use
- Yes
- Sleeps / expires
- No — always on
- Choose a region?
- Auto (global edge, no picker)
- Bandwidth
- Unlimited
- Build Min
- 500
GitHub Pages
checked 2026-09-02- Credit card
- No
- Commercial use
- Yes
- Sleeps / expires
- No sleep
- Choose a region?
- Auto (global CDN, no picker)
- Bandwidth
- 100 GB
- Build Min
- via Actions
Deno Deploy
checked 2026-09-02- Credit card
- No
- Commercial use
- Yes
- Sleeps / expires
- No sleep
- Choose a region?
- Auto (global edge, no picker)
- Bandwidth
- 100 GB
- Serverless
- 1M req/mo
Surge.sh
checked 2026-09-02- Credit card
- No
- Commercial use
- Yes
- Sleeps / expires
- No sleep
- Choose a region?
- Auto (global CDN, no picker)
- Bandwidth
- Unlimited
- Serverless
- None
Cursor
checked- Free tier
- Ongoing Free plan with limited Agent and Chat usage
- Paid from
- Plus $8/month; Premium $30/month
- Model access
- Third-party models require Premium or higher and separate prepaid credits
Cline
checked 2026-09-01- License
- Apache 2.0
- Paid from
- Free (BYOK)
- Shape
- Agent
- Environment
- VS Code
- Setup
- 2 min
Railway
checked 2026-09-14- Free tier
- $0/month, $1 credit included
- Paid from
- ~$5/mo + usage
- Model
- Managed PaaS
- Strengths
- Zero ops, instant, great DX
- Watch-outs
- Costs scale; you don’t own infra
n8n
checked 2026-09-01- Paid from
- Free (self-hosted)
- Best for
- Developers, privacy-conscious teams
- Executions limit
- Unlimited
- Integrations
- 400+ native
- Custom code
- Yes (JS + Python)
Aider
checked 2026-09-01- License
- Apache 2.0
- Paid from
- Free (BYOK)
- Shape
- Agent
- Environment
- Terminal
- Setup
- 1 min (pip)
Appwrite Cloud
checked- Monthly active users
- 75,000
- Free projects
- 2
- Per-project resources
- 1 database, 1 bucket, 2 functions
- Inactivity
- Pauses after 1 week; deletes after 90 paused days
AssemblyAI
checked- Free path
- One-time $50 trial credit, not a monthly allowance
- Illustrative capacity
- Up to 185 hours pre-recorded or 333 hours streaming, depending on model
- After credit
- Upgrade and add funds to continue
- Whisper model
- No; uses AssemblyAI models
Browser Use
checked 2026-09-01- License
- MIT
- Best for
- General autonomous browsing in Python
- ChatOllama
- Local models (Llama, Qwen)
- ChatGoogle
- Google Gemini
- ChatGroq
- Groq
Deepgram
checked- Free path
- One-time $200 signup credit
- After credit
- Pay as you go; no standing free allowance
- Expiry
- No fixed expiration stated
- Whisper model
- Whisper Cloud available
Bolt.new
checked 2026-09-01- Free tier
- 1M tokens/mo, 300K/day
- Paid from
- Free (1M tokens/mo limit)
- Surface
- Browser (in-browser runtime)
- Stack scope
- Full-stack web (Node)
- Default model
- Claude Sonnet 4.6
Firecrawl
checked- Free tier
- 1,000 credits/month
- Examples
- Up to 1,000 pages scraped or 500 searches
- Concurrency
- 2 concurrent requests
- Best for
- Managed scraping and crawling for agents
Oracle Cloud
checked 2026-09-14- Free tier
- 2 OCPU / 12 GB ARM + 2 AMD micro VMs
- Sleeps / expires
- Never
- Free Compute
- 2 ARM cores
- RAM
- 12 GB
- Storage
- 200 GB
Hugging Face
checked 2026-09-08- Free tier
- ~$0.10/month credits (thousands of models)
- Best for
- Trying many open models
- Key / card?
- Account, no card
- Headline model
- FLUX, SDXL, SD 3.5
- Free Embedding Model
- BGE / E5 family
Make.com
checked 2026-09-01- Paid from
- From $9/month
- Best for
- Complex visual workflows
- Executions limit
- 1,000 ops/month (free)
- Integrations
- 1,500+
- Custom code
- Limited
Zapier
checked 2026-09-01- Paid from
- From $19.99/month
- Best for
- Non-technical users, breadth
- Executions limit
- 100-750/month (free)
- Integrations
- 6,000+
- Custom code
- Yes (JS only, paid)
Grok (xAI)
checked 2026-10-11- Free API status
- No permanent free tier; the historical data-sharing credit is not currently documented
- Credit card
- Confirm in your console
- Flagship context
- 1M (the 2M fast models were retired)
- Live search
- Web and X tools available with paid API access
Jina Reader
checked 2026-09-01- Free tier
- Free, generous limit
- Best for
- Dead-simple single-URL reads
- JS render
- Yes
- Structured extract
- Limited
- Self-host
- No (hosted)
ScrapingBee
checked 2026-09-01- Free tier
- 1,000 credit trial
- Best for
- Traditional scraping at scale
- JS render
- Yes
- Structured extract
- Via AI query
- Self-host
- No
Skyvern
checked 2026-09-01- License
- AGPL-3.0
- Best for
- Complex forms across heterogeneous portals
- Language
- Python
- Approach
- Computer-vision agent for form-heavy sites
- WebVoyager
- ~86%
Stagehand
checked 2026-09-01- License
- MIT
- Best for
- TS shops wanting predictable, cheaper runs
- Language
- TypeScript / JS
- Approach
- Hybrid: act/extract/observe + deterministic Playwright
- WebVoyager
- Strong (Browserbase-tuned)
Copilot
checked 2026-09-01- Paid from
- $10/mo
- Surface
- Editor plugin
- Open source
- No
- Choose model
- Mostly OpenAI
HeyGen
checked 2026-08-31- Best for
- Tutorials, training, marketing
- Primary use case
- Talking-head avatars
- Text-to-video
- No (script-to-avatar)
- Image-to-video
- No
Kling
checked 2026-08-31- Best for
- B-roll, hero shots
- Primary use case
- Cinematic clips
- Text-to-video
- Yes (best of three)
- Image-to-video
- Yes (best in class)
Pika
checked 2026-08-31- Best for
- Effects, lip-sync, edits
- Primary use case
- Scene editing & effects
- Text-to-video
- Yes
- Image-to-video
- Yes
Chroma (self-hosted)
checked- Storage
- Your own disk
- Vendor quota
- None for self-hosted software
- Deployment
- Local or your own server
- Cloud option
- Separate usage-based Chroma Cloud plan
Letta
checked 2026-09-01- Credit card
- No
- What it is
- Full stateful-agent runtime
- Core data model
- Editable in-context memory blocks
- License (self-host)
- Apache 2.0
Lovable
checked 2026-09-01- Free tier
- 5 messages/day
- Surface
- Browser (remote sandbox)
- Stack scope
- Full-stack web
- Default model
- Claude / GPT (managed)
Mem0
checked 2026-09-01- Credit card
- No
- What it is
- Drop-in memory layer
- Core data model
- Vector + graph memories
- License (self-host)
- Apache 2.0
Pinecone
checked- Storage
- 2 GB
- Indexes
- Up to 5
- Read units
- Up to 1M/month
- Write units
- Up to 2M/month
Qdrant
checked- Cloud cluster
- 1 node; 0.5 vCPU, 1 GB RAM, 4 GB disk
- Inactivity
- Suspends after 1 week; deletes after 4 weeks if not reactivated
- Best for
- Managed prototype with an open-source self-host path
Replit Agent
checked- Free tier
- Starter plan with daily Agent credits
- Limit
- Limited Agent intelligence; publish 1 app
Zep
checked 2026-09-01- Credit card
- No
- What it is
- Context platform on a temporal graph
- Core data model
- Temporal knowledge graph
- License (self-host)
- Graphiti: Apache 2.0
v0
checked- Free tier
- 7 messages/day
- Deploy
- Can deploy apps to Vercel
- Paid from
- Plus $30/user/month
LiteLLM
checked 2026-09-01- Best for
- Owning your stack, mixing free tiers & local
- What it is
- Open-source library / self-hosted gateway
- Who holds keys
- You — your own provider keys (BYOK)
- Cost model
- $0 software; pay providers directly
bolt.diy
checked 2026-10-24- Paid from
- Free + your own API costs
- Model
- Any model — bring your own key
- Free model option
- Yes — Gemini or Groq, subject to limits
- One-click deploy
- Manual
Azure (B1s)
checked 2026-09-01- Sleeps / expires
- 12 months
- Free Compute
- 1 × B1s
- RAM
- 1 GB
- Storage
- 64 GB SSD
Beam
checked- Free compute
- No standing free allowance on the current pricing page
- Developer plan
- $20/month with $20 in credits
- Concurrency
- Up to 30 CPU or 5 GPU containers
Botpress
checked 2026-08-31- Paid from
- Free tier
- Self-Hosted
- Limited
- No-Code UI
- Yes
- RAG Support
- Partial
Codeium
checked 2026-09-01- License
- Proprietary (free tier)
- Shape
- Autocomplete + chat
- Environment
- 40+ editors
- Setup
- 30 sec
Continue.dev
checked 2026-09-01- License
- Apache 2.0
- Shape
- Autocomplete + chat
- Environment
- VS Code / JetBrains
- Setup
- 10 min (config)
FlowiseAI
checked 2026-08-31- Paid from
- Free (OSS)
- Self-Hosted
- Yes
- No-Code UI
- Yes
- RAG Support
- Yes
GitHub Actions
checked- Free compute
- GitHub Free includes 2,000 minutes/month for private repositories; standard hosted runners are free and unlimited for public repositories
- Job duration
- GitHub-hosted runner jobs can run for up to 6 hours
- Scheduled workflows
- In public repositories, schedules are disabled after 60 days without repository activity
- Credit card
- Not needed within the included allowance; without a payment method, excess private-repo usage is blocked
- Best for
- Finite build or automation jobs, not persistent hosting
Google Cloud (e2-micro)
checked 2026-09-01- Sleeps / expires
- Never
- Free Compute
- 1 × e2-micro
- RAM
- 1 GB
- Storage
- 30 GB HDD
Google Cloud Run
checked 2026-09-01- Credit card
- Yes — billing account required
- Free compute
- 2M requests, 180,000 vCPU-sec, 360,000 GiB-sec, 1 GB North America egress per month
- Long / background work?
- Yes, but pinning always-on instances leaves the free tier
- Sleeps or expires?
- Scales to zero; allowance resets monthly
LangFlow
checked 2026-08-31- Paid from
- Free (OSS)
- Self-Hosted
- Yes
- No-Code UI
- Yes
- RAG Support
- Yes
Tabby
checked 2026-09-01- License
- Apache 2.0
- Shape
- Autocomplete
- Environment
- VS Code / JetBrains / Vim
- Setup
- 5 min (Docker)
Voiceflow
checked 2026-08-31- Paid from
- Paid
- Self-Hosted
- No
- No-Code UI
- Yes
- RAG Support
- Partial
AWS (accounts since July 2025)
checked 2026-09-14- Sleeps / expires
- 6 months, or when credits run out
- Free Compute
- Up to $200 in credits, no free instance
Kimi K2
checked 2026-10-24- Free hosted API
- None verified; current Cloudflare Kimi models require paid access
- OpenRouter
- Kimi models in the live catalog have nonzero token prices
- Other access
- Chat app or self-hosted open weights (hardware required)
CrewAI
checked 2026-09-01- Time to first agent
- ~30 min
- Conditional branching
- Limited
- Loops / retry logic
- Basic
Heroku
checked 2026-09-01- Free web hosting
- No free tier
- Free PostgreSQL
- $5/mo
- Static sites
- Not free
AutoGPT
checked 2026-09-01- Time to first agent
- ~15 min (UI) / ~45 min (SDK)
- Conditional branching
- Via autonomous planning
- Loops / retry logic
- Built-in
LangGraph
checked 2026-09-01- Time to first agent
- ~2–4 hrs
- Conditional branching
- Full — first-class
- Loops / retry logic
- Full control
Qwen3-Coder
checked 2026-10-24- OpenRouter API
- No Qwen3-Coder
:freeID in the live model catalog - Open-weights route
- Self-hosting requires your own hardware
Nothing matches that. Try a shorter word.