nimbus.tools
Get an API key

Changelog

Every release of the Nimbus API and the nimbusapi.net site. Reverse-chronological.

We follow semver at a product level: major for breaking API or billing changes, minor for new models or features, patch for pricing corrections and bug fixes. Pricing corrections that under-charged customers are absorbed by Nimbus; pricing corrections that over-charged customers refund automatically to account balance.

2026-07-01v2.0.0major

Full rewrite of nimbusapi.net

  • New landing page with Three.js WebGL hero animation and cloud-parallax scroll.
  • 28-model catalog surfaced across /docs/pricing and /docs/models with strikethrough retail prices.
  • Prepaid balance model replaces the old subscription tiers. No seat licenses, no commit.
  • Session-cookie auth replaces bearer-token portal auth. API keys still bearer.
  • Docs section rebuilt: 7 new billing/resource pages under /docs/*, real content top to bottom.
  • Admin console at /dashboard/admin gets card grid for users, invoices, refunds, cost caps.
2026-06-30v1.5.0minor

SpiderSense provider + Azure fallback chain

  • SpiderSense (China lab) added as a wholesale provider behind DeepSeek and Qwen models.
  • Azure fallback wired for Anthropic + OpenAI: when the direct upstream returns 5xx or 429, the request is retried against Azure Model Router before returning to the caller.
  • Fallback adds ~120ms p50 latency in the failure path; zero-cost on the success path.
  • New response header x-nimbus-upstream-route echoes which upstream served the request.
2026-06-24v1.4.2patch

DeepSeek V4 Pro added

  • DeepSeek V4 Pro live at $0.215 input / $0.435 output per 1M — 50% off the vendor list of $0.43 / $0.87.
  • 1M context, strong tool calling, competitive with Claude Sonnet on public code benchmarks.
  • Available as deepseek/deepseek-v4-pro on the OpenAI-compatible endpoint.
2026-06-20v1.4.1patch

Gemini 3 Flash Preview pricing corrected

  • Gemini 3 Flash Preview was mispriced at $0.50 / $3.00 per 1M (the retail rate). Corrected to $0.25 / $1.50 (50% off retail).
  • Affected customers received an automatic balance credit for the delta of any usage from 2026-06-15 through 2026-06-20. Notification emails sent 2026-06-20 09:00 UTC.
  • Deployed a regression test that verifies every model in the catalog charges exactly retail ÷ 2.
2026-06-15v1.4.0minor

Claude Opus 4.8 launch

  • Anthropic released Claude Opus 4.8 at 12:00 UTC. Nimbus enabled it in production at 12:47 UTC.
  • Priced at $2.50 input / $12.50 output per 1M (retail: $5.00 / $25.00).
  • 1M context window, extended thinking supported, tool use supported, vision supported.
  • Registered as anthropic/claude-opus-4.8. The 4.7 and 4.6 remain available at the same price.
2026-06-01v1.3.0minor

OpenCode + Continue integration guides

  • Published /docs/opencode with a copy-paste opencode.json config that routes to Nimbus.
  • Published /docs/continue with a config.json snippet for Continue.dev users.
  • Both providers verified to work with streaming, tool calling, and cache-control blocks.
2026-05-15v1.2.0minor

Cost caps per API key

  • Daily and monthly USD spend caps can now be set per API key from the dashboard.
  • Hard-cap mode returns 402 insufficient_balance the moment the cap is hit — no upstream call, zero tokens billed.
  • Soft-cap mode continues to serve but fires alert emails every 5 minutes until raised or reset.
  • Alert thresholds default to 50%, 80%, 95% of the cap. Emails + Discord webhook.
  • See /docs/cost-caps for the full walkthrough.
2026-05-01v1.1.0minor

Multi-provider fallback chains

  • Each API key can now register an ordered fallback chain of model IDs.
  • When the primary returns 429 or 5xx after one internal retry, the gateway automatically tries the next model in the chain.
  • Response header x-nimbus-model-used shows which model actually served the request.
  • Chain is only invoked on transient failures — 4xx from your side (bad request, cap exceeded) never falls through.
2026-04-01v1.0.0major

Initial launch with 12 models

  • Nimbus API launched publicly. OpenAI-compatible endpoint at https://llm.nimbusapi.net/v1.
  • Launch catalog: 3 Claude models, 4 GPT models, 2 Gemini models, DeepSeek V3, Qwen 2.5, Kimi K2.
  • Every price fixed at 50% off the vendor list rate.
  • Prepaid balance model, card + crypto top-up, 30-day refund policy.

Subscribe

The changelog is the source of truth. To get notified when a new model launches or when pricing changes, subscribe to release emails in /dashboard/settings or watch /docs for updates. RSS feed available at /docs/changelog.rss.