12 October 2026 · Developer Tools & Infrastructure · 6 min read
I went through every developer-tool pricing announcement I could verify from the first two weeks of October and counted the direction each one moved. Seven made things worse for developers. One made things better. None of them got the kind of coverage a flagship LLM price cut gets, because none of them are a flagship LLM — they're the smaller, boring line items that quietly eat a budget while nobody's watching the headline models. Every number below I pulled from the vendor's own docs, pricing page, or forum announcement, not from a secondhand aggregator.
| Proveedor | What changed | Before → After | Effective |
|---|---|---|---|
| Groq | gpt-oss-safeguard-20b free-tier rate limit | 30 RPM / 8,000 TPM → 3 RPM / 2,000 TPM | ~Oct 6 |
| volar.io | Machine RAM price | ~$5/GB/mo → $6.00/GB/mo | Oct 2026 |
| Juntos IA | Minimum first credit purchase | $5 → $15 (later purchases stay $5) | Oct 7 |
| exa | One-time sign-up credit | $20 → $10 | Oct 1 |
| OVHcloud | Local storage + IPv4 on Gen3 Public Cloud | Bundled in instance price → billed separately (+1.4% to +21.9%) | Oct 1 |
| Mockoon | Cloud plans | Discontinued, replaced by self-hosted Pro | Oct 1 |
| Oracle Apiary | Entire product | Shut down, all instances deleted | Oct 31 |
| Neón | Free-plan Postgres storage per project | 0.5GB → 1GB (new 20GB account cap) | Oct 1 |
Groq's cut is the sharpest by percentage and the easiest to miss, because it's scoped to a single model. If your app calls gpt-oss-safeguard-20b — Groq's safety-classifier model, not one of the mainline Llama or GPT-OSS chat models — on the free tier, you went from 30 requests a minute to 3 overnight. That's not a gentle throttle, it's a 10x wall. The daily caps (1,000 requests, 200,000 tokens) didn't move, so if you were already pacing yourself to stay under the daily limit you might not notice. If you were leaning on the per-minute headroom to burst through a batch job, you'll hit 429s you've never seen from that endpoint before.
Fly.io's RAM bump from roughly $5 to $6 per GB a month doesn't sound dramatic until you remember RAM is usually the bigger line item than vCPU on a lot of small Machines — a shared-cpu-2x preset reportedly went from $3.89 to $4.39/month, about 13% up for doing nothing differently. OVHcloud's change is structurally sneakier: nothing about the headline instance price moved, but local storage and public IPv4 — both previously baked into the Gen3 instance rate — are now itemized separately starting October 1, and because they also sit outside OVH's savings-plan discount, the real bill moves between +1.4% and +21.9% depending on the instance size, hitting small instances hardest. Neither of these shows up if you're pricing from memory of what you paid last quarter.
Together AI's change doesn't touch per-token pricing at all — it changed the minimum first credit purchase from $5 to $15, per their own billing docs as of October 7. Nobody who's already bought credits is affected. But if you're a student or hobbyist who hasn't signed up yet and was planning to put in $5 to kick the tires, the real entry cost just tripled. Exa did something similar from the other direction: its one-time sign-up bonus dropped from $20 to $10 as of October 1 (the recurring $10/month credit is unchanged), which quietly halves the free runway a brand-new account gets before needing a card.
Mockoon killed its Cloud plans entirely on October 1 and replaced them with self-hosted Mockoon Pro. On price alone this one is a net positive for teams willing to run it themselves: the old Team Cloud plan was $100/month billed annually for 5 seats and capped at 100,000 API calls a month. The new self-hosted Team plan is €10 per license per month (€120/year per license) — a 5-person team runs about €600/year instead of $1,200/year, and since calls never leave your own infrastructure, that 100k/month ceiling is gone too. The catch, which every "self-hosted is cheaper" comparison glosses over, is that you're now the one running and patching it.
Oracle's Apiary (API Blueprint documentation tooling, acquired years ago) reaches full end of life on October 31, 2026. Every instance gets deleted, not archived. If you've still got API docs living on Apiary and haven't migrated, that's a hard deadline, not a renewal notice — Oracle's own guidance points toward OpenAPI-based tools like Swagger Editor, Stoplight or Redocly as the landing spot.
Buried among seven worse deals: Neon doubled its free Postgres plan's per-project storage from 0.5GB to 1GB on October 1, and added a 20GB cap across all projects on an account (previously there wasn't a stated account-wide ceiling at all). Existing projects picked the new limit up automatically — no migration, no action needed. Small, but it's the only item on this list where the number moved in the developer's favor.
None of these eight changes are individually catastrophic. What they have in common is that none of them announce themselves loudly — there's no "price hike" email most of the time, just an updated docs page or a pricing table that quietly swapped one number for another. If your stack touches any of Groq, Fly.io, Together AI, Exa, OVHcloud, Mockoon or Apiary, it's worth five minutes checking whether this month's version of "normal" still matches what you budgeted for last quarter.
¿Eres nuevo en los precios de API basados en el uso? Empieza con el guías gratuitas de costos de API.
Despliéguelo usted mismo: DigitalOcean — $200 de crédito gratis ↗ · VPS Hostinger ↗
Sourced directly from each vendor: Groq's rate-limits docs, Fly.io's pricing page, Together AI's billing-credits docs, Exa's pricing page, OVHcloud's official blog post on the October 2026 Public Cloud pricing update, Mockoon's pricing page, Oracle's service-change announcement for Apiary, and Neon's October 2 changelog. Checked 2026-10-12. Reference figures, not live quotes — confirm current terms directly with each provider before budgeting.