Fully restored

In just 24 hours from yesterday to today we processed over 1 billion tokens in total

Added gemini 3.1 pro and 3.5 flash using our own antigravity pool (we don't currently have a lot, might only last a bit, i might get more later)

Restored gpt 5.6 performance to full speed (it was bad for 20 mins while I figured something out)

Resolved gpt issues, our requests to our partner were suddenly being blocked by cloudflare in the middle of the night for no reason so I connected a backup provider

Everything stable with no issues since i woke up

Identified and fixed weird problems with openai useragents, this means everyone using codex or pi should see significantly faster speeds

How to earn free money 2026 working ^

Our first uncensored model

Fixed caching related problems found on certain harnesses, your balance should be able to go much further now while coding

Updated docs to include windows instructions

Had a 10 mins downtime while i was sleeping but it resolved itself, just posting here to remind everyone I have an EXTRA stable system that most of the time doesn't even need me to be online

*Also fixed $150 of underpaid referral rewards

Reposting official group links
https://t.me/codeifycc
https://discord.gg/KXc5pnuq9d

Gemma 4, not a good coding model but useful if you need to batch process images. Currently testing serving it with cerebras at 1000tps

Introducing Krenn 1 and 2 Mini

These are both built on 12B models, 1 is tuned on Opus 4.8 and 2 is tuned on Fable 5.

Usage will be noticeably slow as these are experimental and I'm just making them public so you could have a go if you were interested

Cheers 🥂

Added composer 2.5 and fixed some toolcalling issues with haiku

Expanded provision for gpt oss 120b, gemma 4 31b, and glm 4.7.
I'll look into adding kimi k3 asap