Blog

Arete 27B is Nous Research's free fine-tune of Qwen3.8-27B. A 52.8M-token training run cut party-line answers from 22% to 0.9% without losing coding scores.

Inkling 2 Small more than doubles Inkling-Small on independent tests and cuts its hallucination rate from 61% to 27%. It is also slower and about 6x pricier per task.

My Inkling 2 Small review: 594B parameters, a 31 on Artificial Analysis versus 13 for Inkling-Small, and a 73.3% non-hallucination rate. What changed and who should care.

Choosing an AI chatbot platform is overwhelming. After testing 9 of the most popular options, I found a clear winner that lets you go live in minutes, not months, by plugging directly into your existing help desk.

Runway ML is a leader in AI video, but is it the best tool for you? Compare 8 Runway AI alternatives for quality, ease of use, and budget in 2026.

Front conversation custom fields add Category, External app ID or Last assignee to a thread. Here is how to create them, the seven types, the limits and what each plan unlocks.

Front can force teammates to fill in conversation custom fields before they archive. Here is how to set it up, the plan it needs, and what quietly skips it.

Front ticket statuses sit on top of archive and snooze: Archived becomes Resolved, Snoozed becomes Waiting. Here is how custom statuses, rules and replies behave.

Front message template variables fill in the contact's name, account fields and sender details. Here is the full list, the fallback syntax and what each plan allows.