Every free tier we have checked, in one place — 89 tools, each with the date we last confirmed its figures. The numbers come from our own articles, and each card links to the one it came from.
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.
Render
checked 2026-09-14- Free tier
- 750 instance hours/month
- Credit card
- No
- Commercial use
- Yes
- Sleeps / expires
- Yes, after 15 min
- Paid from
- Free tier + paid
- Best for
- Full-stack apps, APIs, cron
- The trade
- Easiest deploy of the five
- The catch
- ~1 minute cold start on the next request
- Model
- Managed PaaS
Groq
checked 2026-09-08- Free tier
- True free tier, no card
- Credit card
- No
- Rate limits
- 30
- Speed
- ~521–524 tok/s
- Best for
- Fast batch, hackathons, side projects
- What the minute budget counts
- The max_tokens you declare
- Free limit
- 8,000 TPM
- Is maxTokens: 8192 safe?
- No — every request 413s
Google Gemini
checked 2026-09-08- Free tier
- True free tier, no card
- Credit card
- No
- Speed
- 141 tok/s
- Best for
- Whole documents — 1,000 pages/file, layout + meaning in one call
- Direct call
- ✅ 200
- Inside the harness
- ❌ empty reply, then 120 s timeouts
- What “free” means
- Free tier on Flash models; limits unpublished
OpenRouter
checked 2026-09-08- Free tier
- Multiple free models
- Credit card
- No
- Speed
- ~46–77 tok/s
- Best for
- Zero-ops access to many models fast
- What the minute budget counts
- Nothing per minute — it meters requests
- Free limit
- 20 RPM; 50 RPD, or 1,000 RPD after $10 lifetime spend
- Is maxTokens: 8192 safe?
- Yes; free models allow 32,768 completion tokens and up
DeepSeek
checked 2026-09-02- Free tier
- Low-cost paid (promos)
- Credit card
- Yes
- Speed
- Fast
- Best for
- Coding, reasoning
- What we hit
- 402 Insufficient Balance — the API now needs a funded balance
- Free API path
- Low-cost paid (promos)
- Card to start (free path)
- Yes
Cloudflare Workers AI
checked 2026-09-08- Free tier
- 10,000 Neurons/day (~450–750 checks)
- Credit card
- No
- Speed
- Edge-fast
- Best for
- OCR at the edge inside a Worker you already run
- Images
- No
- Custom policy
- No
- What “free” means
- 10,000 Neurons/day, recurring
Hugging Face Spaces
checked 2026-09-01- Free tier
- 2 Gradio Spaces on ZeroGPU + unlimited static
- Commercial use
- Allowed
- Sleeps / expires
- Yes — after 48 hours idle
- Best for
- Public demos with a shareable URL
- The catch
- Since 2026 a Gradio or Docker Space needs a paid plan to create; ZeroGPU is the free exception
- What you can create free
- Static Spaces (unlimited) + 2 Gradio Spaces on ZeroGPU
- “Good standing”
- Verified email, account older than 30 days
ElevenLabs
checked 2026-09-02- Free tier
- 10,000 credits/month, no card
- Credit card
- Not required
- Commercial use
- ❌ No — needs Starter ($6/mo)
- Best for
- Narration, audiobooks, character voices, cloning
- Voice quality
- Best in class, plus instant cloning
- Paid rate (cheapest)
- $6/mo (30K credits) Starter
- Free monthly volume
- 10,000 credits (~10 min)
Coolify
checked 2026-09-01- Credit card
- Not for Coolify itself (free software). Your server provider may require one
- Commercial use
- Yes — Apache 2.0, no restrictions
- Sleeps / expires
- No — it’s your always-on server
- Paid from
- $0 — Apache 2.0, no feature gates. You pay only for the server, which can be $0 on an Oracle Always Free VPS
- Resources
- Whatever your server has. Coolify needs min 2 vCPU / 2 GB RAM / 30 GB disk; 4 cores / 8 GB is comfortable
- What you can deploy
- Anything containerizable — Node, Python, Go, PHP, Rust, static sites, Docker Compose — plus 280+ one-click services and managed databases
- What you can’t
- Nothing is platform-gated; you’re limited only by your server’s size. It’s not managed hosting — no one runs the box for you
Cloudflare Workers
checked 2026-09-14- Free tier
- 100,000 requests/day
- Credit card
- No
- Sleeps / expires
- No — nothing to idle
- Best for
- Edge APIs, serverless
- The trade
- 10 ms CPU per request; cannot run a process
- Free compute
- 100,000 requests/day, 128 MB per isolate, 10 ms CPU per request, 5 cron triggers
- Long / background work?
- Partly — I/O wait is free, but cron, queue and alarm runs cap at 15 min
Vercel
checked 2026-09-02- Credit card
- No
- Commercial use
- No
- Sleeps / expires
- Minimal
- Best for
- Frontend, Next.js, serverless
- Free Database
- No (add-on)
- Choose a region?
- Auto edge; one function region in settings
- Bandwidth
- 100 GB
Exa
checked 2026-09-01- Free tier
- $10 credit (~1,000 searches)
- Credit card
- No
- Rate limits
- ~5 req/sec
- Best for
- Similarity / research
- Resets
- One-time credit
- Index type
- Neural embeddings
- Content extraction
- Yes ( contents.text )
Tavily
checked 2026-09-01- Free tier
- 1,000 credits/month
- Credit card
- No
- Rate limits
- ~10 req/sec
- Best for
- RAG agents needing clean text
- Resets
- Monthly
- Index type
- Aggregated + own crawl
- Content extraction
- Yes ( include_raw_content )
Brave (Free)
checked 2026-09-01- Free tier
- $5 credit/month (~1,000 queries) — free tier retired Feb 2026
- Credit card
- Yes (since Feb 2026)
- Rate limits
- Up to 50 req/sec (paid tiers)
- Best for
- Cheapest paid search on an independent index
- Resets
- $5 credit monthly
- Index type
- Independent crawl
- Content extraction
- Snippet only (paid adds it)
PocketBase
checked 2026-09-01- Credit card
- None — no vendor account or signup; you only pay for the server
- Commercial use
- Allowed — MIT license, no fee, no seat or request cap
- Best for
- Solo apps, prototypes, internal tools on one server
- Free quota / limits
- None imposed by PocketBase — CPU, RAM, and disk depend entirely on your own server
- Storage / records
- Bounded only by your disk (SQLite file + uploads live in pb_data )
- Projects / instances
- Unlimited — run as many instances as your box can host
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 2026-09-01- Credit card
- No
- Commercial use
- Allowed on free tier
- Best for
- Full-stack apps with auth + storage
- Database
- PostgreSQL 15
- Storage
- 500 MB
- Compute
- Shared, pauses after 1 week inactivity
Neon
checked 2026-09-14- Credit card
- No
- Commercial use
- Allowed on free tier
- Best for
- Dev branches, AI/serverless workloads
- Database
- PostgreSQL 16
- Storage
- 0.5 GB per project (up to 100 projects)
- Compute
- 0.25 vCPU / 1 GB, auto-suspends after 5 min
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 2026-09-02- Credit card
- No
- Sleeps / expires
- No
- Best for
- Static sites, JAMstack
- Free Database
- No
- Choose a region?
- Auto (global CDN, no picker)
- Bandwidth
- 300 credits/month covering both
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
Cerebras
checked 2026-09-01- Rate limits
- 30
- Speed
- not tested
- What we hit
- Free trial now requires a verified payment method up front
- Free Chat
- $5 trial only (card required); gpt-oss-120b
- Free Reasoning
- Limited
Cursor
checked 2026-09-01- Free tier
- 2-week Pro trial
- Paid from
- $20/mo Pro
- Surface
- Local IDE (forked VS Code)
- Stack scope
- Anything on disk
- Default model
- Cursor-managed
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
Mistral AI
checked 2026-09-02- Free tier
- Rate-limited free models
- Speed
- ~161–179 tok/s
- Direct call
- ✅ 200
- Inside the harness
- ❌ fetch failed
- Model tested
- mistral-small
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
checked 2026-09-01- Free tier
- 75K MAU, pauses if idle
- Best for
- Mobile-first apps, Firebase-style DX
- Database
- MariaDB (document API)
- Deployment
- Docker Compose
- Self-host complexity
- Medium
AssemblyAI
checked 2026-09-02- Free tier
- 185 hrs pre-recorded + 333 hrs streaming, no card
- Best for
- Biggest free allowance; summary/sentiment in one call
- Whisper model
- None — Universal only
- Cheapest paid rate
- $0.21/hr Universal-3.5 Pro · $0.15/hr Universal-2
- Native non-Whisper model
- Universal-3 Pro, Universal-2
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 2026-09-02- Free tier
- $200 signup credit
- Best for
- Production with diarization and SLAs
- Whisper model
- Whisper Cloud (whisper-large)
- Cheapest paid rate
- ~$0.48/hr Whisper · $0.258/hr Nova-3
- Native non-Whisper model
- Nova-3, Nova-2, Flux
Together AI
checked 2026-09-08- Free tier
- No longer free — new accounts are read-only until you deposit
- Speed
- ❌ read-only until you deposit
- What we hit
- Account is in read-only mode until you “make an initial deposit”
- Key / card?
- Deposit required
- Headline model
- FLUX.1 [schnell]
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 2026-09-01- Free tier
- 1,000 credits/mo, no card
- Best for
- Managed scrape + crawl for agents
- Scrape / Crawl / Map
- 1 credit per page
- Search
- 2 credits per 10 results
- Interact (browser actions)
- 2 credits per browser-minute
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 2026-09-01- Free tier
- $150/mo via data sharing
- Credit card
- Sometimes (region/deposit)
- Privacy cost of “free”
- Trains on your data
- Flagship context
- 1M (2M on fast)
- Real-time web/X search
- Yes (Live Search)
NVIDIA NIM
checked 2026-09-01- Credit card
- No
- Speed
- 3.5 s
- Free models
- 102 models in the public catalog
- Free limits
- Credit-based
- Model tested
- nemotron-3-super-120b
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)
GLM
checked 2026-09-02- Speed
- 2.6 s
- Free API path
- 3 free Flash models
- Card to start (free path)
- No
- Open weights
- Partly
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
checked 2026-09-01- Credit card
- No (no account needed)
- Storage
- Your disk
- Indexes / collections
- Unlimited (your file system)
- Reads per month
- Unlimited (CPU-bound)
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 2026-09-01- Credit card
- No
- Storage
- 2 GB
- Indexes / collections
- 5 indexes
- Reads per month
- 2 M read units
Qdrant
checked 2026-09-01- Credit card
- No
- Storage
- 1 GB RAM (~1–3M vectors at 384d)
- Indexes / collections
- Multiple in 1 cluster
- Reads per month
- No hard cap (RAM-bound)
Replit Agent
checked 2026-09-01- Free tier
- Limited trial that expires
- Surface
- Browser (remote VM)
- Stack scope
- Full-stack + DB + cron
- Default model
- Claude / GPT (managed)
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 2026-09-01- Free tier
- Generous chat-only
- Surface
- Browser (UI-only)
- Stack scope
- UI components only
- Default model
- v0-managed
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 2026-09-01- Paid from
- Free + your own API costs
- Model
- Any model — bring your own key
- Free model option
- Yes — Gemini , Groq , Together
- 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 2026-09-01- Credit card
- No
- Free compute
- $30/month credits, 30 concurrent CPU / 5 GPU containers
- Long / background work?
- Yes, but scale-to-zero between requests
- Sleeps or expires?
- Credits reset monthly; apps spin down after each request
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 2026-09-01- Credit card
- No
- Free compute
- 2,000 min/month on a free account; standard runners free without limit on public repos
- Long / background work?
- Yes — up to 6 h per job
- Sleeps or expires?
- Runs then exits; public-repo schedules auto-disable after 60 days of no activity
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 2026-09-01- Free API path
- OpenRouter :free + CF Workers AI
- Card to start (free path)
- No
- Open weights
- Yes (Modified MIT)
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 2026-09-01- Free API path
- OpenRouter :free + self-host
- Card to start (free path)
- No
- Open weights
- Yes (Apache 2.0)
Nothing matches that. Try a shorter word.