LIVE FROM THE GPU MINES CLAUDE-HAIKU-4-5 STALE / CLAUDE-OPUS-4-7 STALE / CLAUDE-SONNET-4-6 STALE / DEEPSEEK-V4-FLASH STALE / DEEPSEEK-V4-PRO STALE / GEMINI-3.1-FLASH-LITE STALE / GEMINI-3.1-FLASH-LITE-PREVIEW STALE / GEMINI-3.1-PRO-PREVIEW STALE / GEMINI-3-FLASH-PREVIEW STALE / MISTRAL-LARGE-LATEST STALE / MISTRAL-MEDIUM-LATEST STALE / MISTRAL-SMALL-LATEST STALE / GPT-5.4 STALE / GPT-5.4-MINI STALE / GPT-5.5 STALE /

We poke 14 models every 2 minutes. No vibes - just 200s and 503s.

pick a model and hit check

SPEED RACE

anthropic/claude-opus-4-7

- STALE

deepseek/deepseek-v4-pro

- STALE

google/gemini-3.1-pro-preview

- STALE

mistral/mistral-large-latest

- STALE

openai/gpt-5.5

- STALE

THE NUMBERS

active incidents

AIs so very weary

avg response time

probes (24h)

wrong per 100

LAST MELTDOWNS

google/gemini-3.1-pro-preview timed out

2 weeks, 3 days ago

google/gemini-3.1-pro-preview timed out

2 weeks, 3 days ago

google/gemini-3.1-pro-preview timed out

2 weeks, 4 days ago

google/gemini-3.1-pro-preview timed out

2 weeks, 4 days ago

HELP US GET MORE

EVERY 2 MIN

5 PROVIDERS

14 MODELS

Help us add GROK and send our 60 SECOND probes, with Llama, Qwen, and more coming after. The more traffic we get, the more tokens we can afford / spread the word.

SHARE ON X SHARE ON LINKEDIN

HOW IT WORKS

WE KICK EVERY MODEL

Every 2 minutes we send a unique prompt to every AI provider. Each one is different. Nothing gets cached.

WE CHECK THE ANSWER

Each prompt has a right answer. If the model gets it wrong, we notice.

YOU SEE WHAT'S REAL

Speed rankings, response times, and which providers are actually down — updated live.