What’s new in RackRadar.

Monthly comparisons now use the recurring price alone. One-time setup charges remain separate in offer details and no longer inflate budget filters, CPU/GPU value calculations, or monthly unit costs. A $2 monthly plan with a $10 setup charge therefore compares at $2 per month. The public API exposes pricing.recurringMonthlyUsd; its existing effectiveMonthlyUsd field remains a compatibility alias. Value methodology version 2 records this change.

Choosing a provider now scopes the offer overview to its published inventory: offer count, lowest recurring monthly price, latest observation, and allocation models. Other filters continue to narrow the results below that overview.

24fire’s EYG location now resolves to Eygelshoven in the Netherlands, with the provider-specific location evidence retained.

Site operators can now review provider requests and offer correction reports in a private dashboard. Search and status filters, dated offer snapshots, source links and required review notes keep the review context together.

Admin access must first be configured locally; there is no default password. Reviews update the private queue only. Published provider coverage and offers continue through the research, collector and import workflow.

Collected coverage now includes Alwyzon, Datalix, 1984 Hosting, Layer7, V.PS, Evoxt, HostSlick, HostUp, Civo and HostDare. Find them under Available to compare in the provider directory, or use the provider filter in Compare offers.

Each collector covers specific published product tables or plan families. Coverage does not include every product or establish current stock. Original currency, billing qualifications and unknown specifications stay visible; Civo's monthly reference is distinguished from its hourly billing.

Public API v1 now includes GET /api/v1/offers/{offerId}/history with bounded cursor pagination. Each observation retains its date, source URL and original price, currency, billing period, setup fee and tax basis.

These are observations recorded when material terms changed, not a continuous price feed or live availability. Raw source payloads and private submissions remain excluded. See the API reference.

The homepage now puts comparison starting points and the provider directory closer together. Shared VPS, dedicated cores and bare metal link straight to their filters; GPU servers link to the separate GPU comparison.

Provider shortcuts are quieter, and coverage counts sit beside the directory. Collected offers and the research backlog remain separate, with the same search, filters and account-free provider request form.

Browser Back also returns to the same place in the directory after visiting offer comparisons.

Provider logo frames now use one white surface and border in both themes, including fallback initials. This removes the mixed green, white and pastel tiles from provider lists, offer details and comparison.

24fire uses its official black-lettered logo; RackNerd and Webdock use official compact symbols. Original logo colors and proportions are preserved, with local assets and recorded sources.

GPU servers now have a dedicated comparison view, separate from CPU compute offers. Filter by exact GPU model, total VRAM, GPU count and access, host allocation, RAM, budget and the location atlas.

The initial collection includes eight published configurations from Hetzner and Vultr. GPU capability stays independent of allocation: a GPU host can still be a cloud instance or bare-metal server. The Hetzner collector also follows its updated GPU matrix and retains the four-GPU default configuration.

GPU offers use VRAM value, allocated GB per $100 of effective monthly cost, with price per GPU beside it. This is a memory-capacity metric, not a benchmark. Hourly prices and monthly references retain their billing distinctions; known setup fees were initially included in the effective comparison cost. The recurring-price update separates these one-time charges under methodology version 2.

Quick view, shareable offer pages, correction reports and the global shortlist support GPU details. The read-only API adds GPU fields and a capability filter. Read how GPU comparisons work for the formula, sources and limitations.

Switching between compute and GPU servers now keeps the tab bar mounted, with a sliding active underline through loading states. The GPU and microchip icons use restrained hover/press feedback, with immediate keyboard and reduced-motion behavior.

Public API documentation and the Changelog are now available in the top navigation and mobile menu, alongside providers, offer comparisons and guides.

Report incorrect information is now an outlined button with a flag icon, grouped with the offer’s comparison and source actions. The same placement appears in quick view and on offer pages, with larger touch targets on phones.

RackRadar’s collected hosting data is now available through a read-only public API. Build your own tools with versioned endpoints for offers, individual offer details, collected providers and reviewed location metadata. No account or API key is needed.

The API documentation includes interactive requests and a downloadable OpenAPI specification. Lists use bounded cursor pagination; offer filters include provider, allocation, kind, observation status and minimum published RAM or cores.

Responses preserve original prices and billing commitments, source links, observation dates and unknown values. Value scores use the same editorial method as the comparison UI. These are collected observations, not live stock or measured performance. Private reports, provider requests and raw source payloads are excluded.

Spot a price that has changed, an incorrect specification or a broken source? You can now report incorrect information from an offer’s quick view or its shareable page, without an account.

Choose what looks wrong and explain what you noticed. You can also suggest a correction or link to evidence. RackRadar attaches the offer and its recorded details, so you don’t need to copy everything yourself.

Reports go into a private review queue. They do not change published offers or value scores automatically. Your draft stays available if you return to the offer before sending it.