Model status

Every model we can currently serve is checked with a real minimal request. The table below is the result of that probe — not a hand-maintained board. Anything that fails shows as degraded rather than being quietly hidden.

Waiting on first probe

The health probe has not returned a reading yet.

We do not publish an uptime SLA. Rather than an invented uptime percentage, each model row below shows the current probe result and, where available, its measured round-trip latency.

ModelModel IDStatusLatencyUptimeLast checked
GPT-5.6 Solopenai/gpt-5.6-solNot probednot probed
GPT-5.6 Terraopenai/gpt-5.6-terraNot probednot probed
GPT-5.6 Lunaopenai/gpt-5.6-lunaNot probednot probed
GPT-5.5openai/gpt-5.5Not probednot probed
GPT-5.5 Proopenai/gpt-5.5-proNot probednot probed
GPT-5.4openai/gpt-5.4Not probednot probed
GPT-5.4 Miniopenai/gpt-5.4-miniNot probednot probed
GPT-5.4 Nanoopenai/gpt-5.4-nanoNot probednot probed
GPT-5.4 Proopenai/gpt-5.4-proNot probednot probed
GPT-5.2openai/gpt-5.2Not probednot probed
GPT-5openai/gpt-5Not probednot probed
GPT-5 Miniopenai/gpt-5-miniNot probednot probed
GPT-5 Nanoopenai/gpt-5-nanoNot probednot probed
Gemini 3.1 Progoogle/gemini-3.1-pro-previewNot probednot probed
Gemini 3.6 Flashgoogle/gemini-3.6-flashNot probednot probed
Gemini 3.5 Flashgoogle/gemini-3.5-flashNot probednot probed
Gemini 3 Flashgoogle/gemini-3-flash-previewNot probednot probed
Gemini 3.1 Flash Litegoogle/gemini-3.1-flash-liteNot probednot probed
Gemini 2.5 Progoogle/gemini-2.5-proNot probednot probed
Gemini 2.5 Flashgoogle/gemini-2.5-flashNot probednot probed
Gemini 2.5 Flash Litegoogle/gemini-2.5-flash-liteNot probednot probed
GPT-OSS 120Bgroq/openai/gpt-oss-120bNot probednot probed
GPT-OSS 20Bgroq/openai/gpt-oss-20bNot probednot probed
Qwen 3.6 27Bgroq/qwen/qwen3.6-27bNot probednot probed
Groq Compoundgroq/groq/compoundNot probednot probed
Groq Compound Minigroq/groq/compound-miniNot probednot probed
Mistral Largemistral/mistral-large-latestNot probednot probed
Mistral Mediummistral/mistral-medium-latestNot probednot probed
Mistral Smallmistral/mistral-small-latestNot probednot probed
Magistral Smallmistral/magistral-small-latestNot probednot probed
Codestralmistral/codestral-latestNot probednot probed
Devstral Mediummistral/devstral-medium-latestNot probednot probed
Ministral 8Bmistral/ministral-8b-latestNot probednot probed
Ministral 3Bmistral/ministral-3b-latestNot probednot probed
Nano Banana Progoogle/gemini-3-pro-imageNot probednot probed
Nano Banana 2google/gemini-3.1-flash-imageNot probednot probed
Nano Bananagoogle/gemini-2.5-flash-imageNot probednot probed
GPT Image 2openai/gpt-image-2Not probednot probed
GPT Image 1 Miniopenai/gpt-image-1-miniNot probednot probed
Nova Voiceopenai/gpt-4o-mini-ttsNot probednot probed
Gemini Voice Flashgoogle/gemini-2.5-flash-ttsNot probednot probed
Gemini Voice Progoogle/gemini-2.5-pro-ttsNot probednot probed
Gemini Voice Litegoogle/gemini-2.5-flash-lite-preview-ttsNot probednot probed
Gemini Voice 3.1google/gemini-3.1-flash-tts-previewNot probednot probed
Nova Transcribeopenai/gpt-4o-transcribeNot probednot probed
Nova Transcribe Miniopenai/gpt-4o-mini-transcribeNot probednot probed
Veo 3.1 Litegoogle/veo-3.1-liteNot probednot probed
Veo 3.1 Fastgoogle/veo-3.1-fastNot probednot probed
Veo 3.1google/veo-3.1Not probednot probed

Listed, not yet servable

These providers are fully integrated, but their accounts are awaiting funding from the NovaServe team. They are not probed and any request for them is refused with a clear error — never rerouted to a different model.

  • Claude Opus 4.5anthropic/claude-opus-4-5Awaiting funding
  • Claude Sonnet 4.5anthropic/claude-sonnet-4-5Awaiting funding
  • Claude Haiku 4.5anthropic/claude-haiku-4-5Awaiting funding
  • Grok 4.1xai/grok-4.1Awaiting funding
  • Grok 4.1 Fastxai/grok-4.1-fastAwaiting funding
  • Grok Code Fastxai/grok-code-fast-1Awaiting funding
  • DeepSeek V3.2 Chatdeepseek/deepseek-chatAwaiting funding
  • DeepSeek Reasonerdeepseek/deepseek-reasonerAwaiting funding

Incident history

No incidents in the last 30 days.

How these numbers are produced

  • Each model receives a small live request through the same path your API calls take.
  • Latency is the measured round trip of that probe, not a published figure.
  • Results are cached briefly, so a reading can be up to a few minutes old.
  • We do not publish an uptime percentage or an SLA; what you see here is the current probe.