kontinent / status
Measured model availability
We call every served model every 30 minutes with the same prompt, through our own gateway. What you see here is the result of those calls — not a third-party report and not an estimate.
At least one provider is failing
As of 8/28/26, 1:21 PM UTC · measuring since Aug 19, 2026
ProviderAvailableResponse
AKI.IO96.7 %724 ms
Qwen3.8 27B (AKI.IO)66.7 %918 ms
Apertus 70B100.0 %1.2 s
DeepSeek V4 Flash (AKI.IO)100.0 %1.0 s
Gemma 4 26B (AKI.IO)100.0 %715 ms
GPT-OSS 120B (AKI.IO)100.0 %258 ms
Kimi K2.7-Code 1100B100.0 %612 ms
Llama 3.1 8B100.0 %269 ms
Llama 3.3 70B (AKI.IO)100.0 %262 ms
MiniMax M2.5 230B100.0 %412 ms
Mistral 4 119B100.0 %549 ms
Qwen3.6 35B (AKI.IO)100.0 %735 ms
AWS Bedrock (EU Frankfurt)100.0 %1.1 s
Claude Haiku 4.5 (Bedrock)100.0 %894 ms
Claude Opus 4.5 (Bedrock)100.0 %1.5 s
Claude Opus 4.6 (Bedrock)100.0 %1.3 s
Claude Opus 4.7 (Bedrock)100.0 %1.2 s
Claude Opus 4.8 (Bedrock)100.0 %1.1 s
Claude Opus 5 (Bedrock)100.0 %2.1 s
Claude Sonnet 4.5 (Bedrock)100.0 %1.9 s
Claude Sonnet 4.6 (Bedrock)100.0 %1.1 s
Claude Sonnet 5 (Bedrock)100.0 %2.5 s
Devstral 2 123B (Bedrock)100.0 %312 ms
GLM 4.7 Flash (Bedrock)100.0 %326 ms
GPT-OSS 120B (Bedrock)100.0 %462 ms
GPT-OSS 20B (Bedrock)100.0 %390 ms
MiniMax M2.1 (Bedrock)100.0 %811 ms
MiniMax M2.5 (Bedrock)100.0 %570 ms
Nova 2 Lite (Bedrock)100.0 %494 ms
Nova Lite (Bedrock)100.0 %580 ms
Nova Micro (Bedrock)100.0 %475 ms
Nova Pro (Bedrock)100.0 %1.2 s
Pixtral Large (25.02) (Bedrock)100.0 %830 ms
Qwen3 235B A22B 2507 (Bedrock)100.0 %337 ms
Qwen3 32B (dense) (Bedrock)100.0 %289 ms
Qwen3-Coder-30B-A3B-Instruct (Bedrock)100.0 %266 ms
AWS Bedrock (EU Irland)100.0 %486 ms
Gemma 3 12B (Bedrock)100.0 %390 ms
Gemma 3 27B (Bedrock)100.0 %423 ms
Gemma 3 4B (Bedrock)100.0 %375 ms
Magistral Small (Bedrock)100.0 %509 ms
MiniMax M2 (Bedrock)100.0 %2.2 s
Ministral 3 14B (Bedrock)100.0 %615 ms
Ministral 3 3B (Bedrock)100.0 %158 ms
Ministral 3 8B (Bedrock)100.0 %384 ms
NVIDIA Nemotron Nano 12B v2 (Bedrock)100.0 %395 ms
NVIDIA Nemotron Nano 3 30B (Bedrock)100.0 %381 ms
NVIDIA Nemotron Nano 9B v2 (Bedrock)100.0 %2.0 s
Qwen3 Next 80B A3B (Bedrock)100.0 %339 ms
Qwen3 Next 80B Thinking (Nebius)100.0 %491 ms
Qwen3-VL 235B A22B (Bedrock)100.0 %465 ms
Azure OpenAI (EU Data Zone)100.0 %1.0 s
GPT-4.1 (Azure EU)100.0 %913 ms
GPT-4.1 mini (Azure EU)100.0 %526 ms
GPT-4.1 nano (Azure EU)100.0 %389 ms
GPT-4o mini (Azure EU)100.0 %1.0 s
GPT-5 (Azure EU)100.0 %3.1 s
GPT-5 mini (Azure EU)100.0 %1.1 s
GPT-5 nano (Azure EU)100.0 %999 ms
GPT-5.1 (Azure EU)100.0 %1.1 s
GPT-5.4 (Azure EU)100.0 %947 ms
GPT-5.4 mini (Azure EU)100.0 %579 ms
GPT-5.5 (Azure EU)100.0 %1.0 s
GPT-5.6 Luna (Azure EU)100.0 %640 ms
GPT-5.6 Sol (Azure EU)100.0 %1.5 s
GPT-5.6 Terra (Azure EU)100.0 %556 ms
o3 (Azure EU)100.0 %1.8 s
o4-mini (Azure EU)100.0 %1.2 s
Google Vertex AI (EU)88.5 %942 ms
Gemini 2.5 Pro (Vertex)0.0 %—
Gemini 2.5 Flash (Vertex)100.0 %843 ms
Gemini 2.5 Flash-Lite (Vertex)100.0 %489 ms
Gemini 3.1 Flash-Lite (Vertex)100.0 %566 ms
Gemini 3.5 Flash (Vertex)100.0 %1.0 s
Gemini 3.5 Flash-Lite (Vertex)100.0 %591 ms
Gemini 3.6 Flash (Vertex)100.0 %1.3 s
Gemini 3.7 Flash (Vertex)100.0 %2.6 s
Inceptron100.0 %1.4 s
DeepSeek V4 Flash (AKI.IO)100.0 %10.1 s
DeepSeek V4 Flash (Inceptron)100.0 %375 ms
GLM 5.2 (Inceptron)100.0 %1.6 s
Kimi K2.6 (Inceptron)100.0 %1.4 s
Kimi K2.7-Code 1100B (Inceptron)100.0 %1.4 s
Infercom100.0 %762 ms
Gemma 4 31B (Infercom)100.0 %2.1 s
GPT-OSS 120B (Infercom)100.0 %215 ms
MiniMax M2.7 (Infercom)100.0 %762 ms
IONOS AI Model Hub100.0 %2.0 s
GPT-OSS 120B (IONOS)100.0 %763 ms
Llama 3.1 405B (IONOS)100.0 %14.9 s
Llama 3.1 8B (IONOS)100.0 %171 ms
Llama 3.3 70B (IONOS)100.0 %333 ms
Mistral NeMo (IONOS)100.0 %873 ms
Mistral Small 24B (IONOS)100.0 %601 ms
Qwen3 Coder Next (IONOS)100.0 %533 ms
Qwen3.5 397B (IONOS)100.0 %3.2 s
Qwen3.5 9B (IONOS)100.0 %3.0 s
Mistral La Plateforme100.0 %496 ms
Codestral100.0 %536 ms
Devstral 2100.0 %281 ms
Ministral 3 14B100.0 %455 ms
Ministral 3 3B100.0 %280 ms
Ministral 3 8B100.0 %997 ms
Mistral Large 3100.0 %113.8 s
Mistral Medium 3.5100.0 %345 ms
Mistral Small 4100.0 %673 ms
Nebius Token Factory100.0 %1.9 s
Cosmos3 Super Reasoner (Nebius)100.0 %2.3 s
Gemma 3 27B (Nebius)100.0 %1.1 s
GLM-5.1 (Nebius)100.0 %3.8 s
GPT-OSS 120B (Nebius)100.0 %200 ms
Hermes 4 405B (Nebius)100.0 %289 ms
Hermes 4 70B (Nebius)100.0 %778 ms
Kimi K3 (Nebius)100.0 %107.2 s
Llama 3.3 70B (Nebius)100.0 %13.1 s
MiniCPM-V 4.5 (Nebius)100.0 %329 ms
Nemotron 3 Nano 30B (Nebius)100.0 %317 ms
Nemotron 3 Nano Omni (Nebius)100.0 %184 ms
Nemotron Ultra 253B (Nebius)100.0 %2.3 s
Qwen2.5-VL 72B (Nebius)100.0 %632 ms
Qwen3 235B A22B (Nebius)100.0 %2.0 s
Qwen3 30B A3B (Nebius)100.0 %110 ms
Qwen3 32B (Nebius)100.0 %4.1 s
Qwen3 Next 80B Thinking (Nebius)100.0 %1.8 s
Nebul100.0 %703 ms
GLM-5.1 (Nebul)100.0 %501 ms
GLM-5.2 (Nebul)100.0 %703 ms
GPT-OSS 120B (Nebul)100.0 %318 ms
Ministral 3 14B (Nebul)100.0 %166 ms
Mistral Large 3 (Nebul)100.0 %590 ms
Nemotron 3 Super 120B (Nebul)100.0 %1.7 s
Qwen3 30B A3B (Nebul)100.0 %445 ms
Qwen3-VL 235B Thinking (Nebul)100.0 %1.3 s
Qwen3.5 397B (Nebul)100.0 %1.3 s
OVHcloud AI Endpoints100.0 %1.6 s
GPT-OSS 120B (AKI.IO)100.0 %6.7 s
GPT-OSS 120B (OVHcloud)100.0 %738 ms
GPT-OSS 20B100.0 %697 ms
Llama 3.3 70B (OVHcloud)100.0 %421 ms
Mistral 7B v0.3 (OVHcloud)100.0 %392 ms
Mistral NeMo (OVHcloud)100.0 %443 ms
Mistral Small 3.2 (OVHcloud)100.0 %5.2 s
Qwen2.5 VL 72B100.0 %254 ms
Qwen3 32B (OVHcloud)100.0 %2.9 s
Qwen3 Coder 30B (OVHcloud)100.0 %535 ms
Qwen3.5 397B (OVHcloud)100.0 %1.8 s
Qwen3.5 9B100.0 %4.3 s
Qwen3.6 27B100.0 %1.7 s
Regolo AI100.0 %853 ms
Apertus 70B (Regolo)100.0 %574 ms
Gemma 4 31B (Regolo)100.0 %147 ms
GLM 5.2 (Regolo)100.0 %2.9 s
GPT-OSS 120B (Regolo)100.0 %256 ms
GPT-OSS 20B (Regolo)100.0 %205 ms
Llama 3.3 70B (Regolo)100.0 %259 ms
Mistral Small 4 (Regolo)100.0 %101 ms
Qwen3 Coder Next100.0 %128 ms
Qwen3.5 122B (Regolo)100.0 %828 ms
Qwen3.5 9B (Regolo)100.0 %969 ms
Qwen3.6 27B (Regolo)100.0 %2.7 s
Qwen3.8 27B (Regolo)100.0 %4.8 s
Scaleway Generative APIs100.0 %967 ms
DeepSeek V4 Flash (Scaleway)100.0 %453 ms
Gemma 4 26B100.0 %641 ms
Gemma 4 26B (AKI.IO)100.0 %13.2 s
GLM 5.2100.0 %1.1 s
GPT-OSS 120B100.0 %294 ms
Llama 3.3 70B (Scaleway)100.0 %99 ms
Mistral Medium 3.5 (Scaleway)100.0 %120 ms
Mistral Small 3.2 (Scaleway)100.0 %154 ms
Qwen3 235B100.0 %894 ms
Qwen3 235B (TensorX)100.0 %10.2 s
Qwen3.5 397B100.0 %2.3 s
Qwen3.5 397B (OVHcloud)100.0 %8.7 s
Qwen3.6 35B100.0 %751 ms
Qwen3.6 35B (AKI.IO)100.0 %5.5 s
TensorX100.0 %1.9 s
DeepSeek R1 05.28100.0 %12.4 s
DeepSeek V3.2100.0 %2.6 s
DeepSeek V4 Flash (0731) (TensorX)100.0 %3.4 s
DeepSeek V4 Pro100.0 %848 ms
GLM 5100.0 %2.4 s
GLM 5 Turbo100.0 %1.0 s
GLM 5.1100.0 %1.4 s
GLM 5.2 (TensorX)100.0 %1.8 s
GLM 5V Turbo100.0 %1.8 s
Kimi K2.5100.0 %2.0 s
Kimi K2.6100.0 %16.0 s
Kimi K2.7-Code 1100B (TensorX)100.0 %405 ms
Kimi K3100.0 %4.6 s
MiniMax M2.5 230B100.0 %6.6 s
MiniMax M2.5 230B (TensorX)100.0 %1.8 s
MiniMax M3100.0 %613 ms
Qwen3 235B (TensorX)100.0 %1.3 s
Qwen3.5 122B100.0 %2.5 s
Qwen3.5 9B (TensorX)100.0 %2.1 s
Qwen3.8 2.4T100.0 %430 ms
Qwen3.8 27B100.0 %826 ms
How we measure
The source is our own probe traffic only, never customer traffic. Every 30 minutes the same short prompt goes to every served model, plus one realistically long prompt once a day. “Available” is the share of successful calls over the last 24 hours; “Response” is the median response time of successful calls. Below 20 samples we show no state at all, because the number would promise more than it carries.
This page describes the availability of individual models and providers. It is not a statement about the platform availability committed in § 7 of the Terms of Service — we expressly do not warrant any particular availability of individual models.