Re-verify provider data with footnoted citations; add 5 new providers
Comprehensive web-research update of the README, verified July 12, 2026 against official docs/pricing pages (cross-checked via official pricing APIs and docs source repos where vendor sites block fetching). Two commits:
Commit 1 — Accuracy pass + footnote citations
Every table cell now cites its source as a GitHub-style footnote (139 footnotes render at the bottom of the page); the References section keeps the broader per-provider link collections, including sources not cited directly.
Corrections found by re-verification:
- Lambda@Edge is metered at 1 ms granularity (README said 50 ms — that changed back in 2021); viewer-event packages cap at 1 MB (50 MB is origin-only); filled the request/response body limits (40 KB viewer / 1 MB origin)
- Azure Functions: Python 3.14 went GA (~early July 2026); max HTTP request is 210 MB (not 100); Flex sizes are 0.5/2/4 GB; deployment package cap is 1 GB; spending limits only exist for credit-based subscriptions
- Koyeb: cheapest paid instance corrected to $0.0000006/s (eco-nano ≈ $1.61/mo) — previous figure was 10× off; Serverless Postgres is GA; joining Mistral AI (Feb 2026) added to platform changes
- Scaleway: Python 3.14 shipped in the July 2026 runtime wave; June 2026 price-change caveat noted
- Render: 2026 plans are Hobby $0 / Pro $25 / Scale $499 with 5 / 25 / 1,000 GB included bandwidth; new-service default is Python 3.14.3 + uv; max standard instance is 8 CPU / 32 GB; community forum retired for Discord
- Tencent: max memory is 3 GB standard with large-memory specs to 120 GB (was "4 GB ❔"); confirmed 24 h async and 500 MB code limits
- Alibaba: filled default timeout (3 s) and memory (console 512 MB); middle CU tier ($0.000017) added
- Bill-limit column made honest across the board: AWS/GCP/OCI/IBM/DO/Scaleway/Huawei budgets are alerts-only, no hard caps; Cloudflare has per-invocation CPU caps only
- Cloud Run: default concurrency is now 80×vCPUs for CLI deploys; 32 MB payload cap is HTTP/1-only
- Vercel: Pro includes $20/mo usage credit (Sept 2025); per-unit invocation billing (May 2026); 1,800 s beta duration; 5 GB "Large Functions" beta
- Modal: $30 credit confirmed monthly; 336 GiB max memory confirmed; Fly.io trial is "2 machine-hours or 7 days"; GPU EOL Aug 1, 2026; Fermyon Wasm Functions GA Nov 2025 → rebranding to "Akamai Functions"
- AWS App Runner exclusion updated: closed to new customers Apr 30, 2026
- Performance table: cold-start cells filled from Cloudflare's published cross-vendor FastAPI benchmark (CF ~1.0 s, Lambda ~2.5 s, Cloud Run ~3.1 s), flagged as vendor-run
Commit 2 — Five new providers added to every table
| Provider | Why it's in |
|---|---|
| Appwrite Functions | Python 3.14 handler FaaS, Cloud GA Aug 2025, 750k free executions/mo (free projects pause after 7 idle days — noted) |
| Azure Container Apps | The Azure analog of Cloud Run: scale-to-zero serverless containers, 180k vCPU-s + 360k GiB-s + 2M reqs/mo free, idle-rate billing, serverless GPUs |
| Huawei Cloud FunctionGraph | Major-cloud FaaS with a permanent 1M reqs + 400k GB-s/mo free tier, Python 3.12, WSGI/ASGI via HTTP functions, 72 h async; no US regions (noted) |
| Railway | Scale-to-zero container platform: $1/mo free credit (since Aug 2025), hard spend caps, per-minute usage billing, S3-compatible Buckets |
| Yandex Cloud Functions | Python 3.14 (unusually current), requirements.txt auto-install, 1M invocations + 10 GB×hr + 100 GB egress free monthly; Russia/Kazakhstan region caveats noted |
Plus a new Watchlist and reviewed-but-excluded subsection: Wasmer Edge, Leapcell, Sevalla, and Blaxel as watchlist entries; documented exclusion rationale for Heroku, Northflank, Zeabur (Apr 2026 pivot), PythonAnywhere, Genezio (pivoted away from FaaS), and Deta Space (shut down); fal added to the GPU-first out-of-scope list; Cloudflare Containers GA (Apr 2026) noted as the full-CPython route on Cloudflare.
Verification notes
- Data gathered by parallel research agents against official sources; where vendor sites block automated fetching, values were confirmed via the AWS Price List API, MicrosoftDocs/azure-docs, cloudflare-docs, Scaleway/Fermyon/Railway/Yandex docs source repos, and official SDK sources
- Huawei's international USD rates are only exposed in its console calculator, so they carry the ❔ marker with China list prices cited for reference
markdownlint(repo config) passes; footnote refs/definitions validated programmatically (139/139, no orphans)
🤖 Generated with Claude Code
https://claude.ai/code/session_01X4AMvv1KYYvnmpkDZeW7Ra
Generated by Claude Code