FIREWORKS

🟡 Turns out the 'server' in serverless was doing all the heavy lifting—now it's on vacation. Multiple models (Ember, Kimi, GLM, Inkling) are limping along while engineers pretend this wasn't completely avoidable.

⏰ 39m ago It's bad spotted via status_rss

// The official story

CSP outage affecting multiple serverless models

// Status page says

Status: Investigating We are experiencing a cloud service provider (CSP) outage affecting multiple serverless models. Requests to affected models may fail. We are investigating the impact. Affected components Ember-1 (Degraded performance) Kimi K3 (Degraded performance) Kimi K3 US (Degraded performance) GLM 5.3 Flash (Degraded performance) Inkling (Degraded performance) GLM 5.3 Fast (Deg

🔗 Official status page (for the PR-approved version)

https://status.fireworks.ai//incidents/01M4YAMVS2260M80QZMQHBPZPY

Fireworks's Greatest Hits

🟢 FIREWORKS Aug 18, 2026

🟠 Kimi K3 is having a regional identity crisis—elevated errors and degraded availability spreading like a bad wifi signal. Engineers are investigating why one region apparently ghosted the network. The mitigation is happening, but 'investigating' means they're still reading the error logs like tea leaves. (Auto-resolved: no updates in 24h)

🟢 FIREWORKS Aug 11, 2026

🟡 Your favorite open-source LLMs are playing dead across the board—GPT-OSS, DeepSeek, Kimi, GLM, the whole squad ghosted. Fireworks is investigating why their serverless infrastructure decided to go full Sleeping Beauty. Apparently 'partial outage' is code for 'we're as confused as you are'. (Auto-resolved: no updates in 24h)

🟢 FIREWORKS Jul 31, 2026

⚪ Update: Qwen3 Embedding 8B embeddings API recovered

🟢 FIREWORKS Jul 31, 2026

🔴 Qwen3's embedding engine decided it was tired of understanding language and just stopped showing up to work. Your vectors are now just vibes in the void. Don't worry, the fix is definitely coming any minute now, probably.