Which LLMs/providers power Qaf?
Category: Questions & membership Priority: P3 Severity: n/a Area: ai-answers Reports: 2 Status: Open
Summary
Two users ask about the tech stack behind Qaf's AI: whether it is built on OpenAI or Claude, and more broadly which LLMs/providers are used, whether Qaf runs its own models, and why. Abdellatif answered both: Qaf uses OpenAI — specifically GPT 5.6 "sol" via the open-source agentset library — and runs no in-house models, noting sol was found to be very good at processing Arabic text. This is a recurring, easily-answered transparency question and a good candidate for a public FAQ entry.
What users say
Adi Alsulti ady19996@icloud.com
- Date: 2026-08-06 (06:01 +04:00)
- Language: English
- Thread:
19fd4ce5eb524613 - Replied by Abdellatif: Yes ("We're using OpenAI.")
Asks — in the subject line only; the message body was empty ("Sent from my iPhone") — whether Qaf is built on OpenAI or Claude.
Which ai do you use? Openai or claude
(Question posed in the subject line; the email body contained only "Sent from my iPhone".)
Abdellatif's reply:
We're using OpenAI.
Mohammad muabt@proton.me
- Date: 2026-08-08 (06:11 +00:00)
- Language: English
- Thread:
19fdfff9d24dbea5 - Replied by Abdellatif: Yes (agentset + GPT 5.6 sol, no own models)
Praises the app, then asks for transparency about the system behind Qaf: which LLMs/providers power the AI agent, whether they run their own models, and whether there is a specific reason for the choice.
I'm curious if you can be more transparent about the system behind Qaf. What LLMs/providers are used for the Qaf AI agent? Or do you run your own models? And is there a specific reason for that choice?
Abdellatif's reply:
Wa alaikum alsalam, we're mostly using the agentset open source library with GPT 5.6 sol. We don't run any of our own models.
Found sol to be really good in processing Arabic text.
JAK for the kind words!
Details / repro
- Answered stack: OpenAI, GPT 5.6 "sol", via the open-source agentset library; no in-house/self-hosted models.
- Stated rationale: sol performs very well on Arabic text.
- Note: Adi's question lives entirely in the subject line ("Which ai do you use? Openai or claude"); the body was empty — worth remembering when triaging subject-only emails.
- Suggested handling: a public FAQ entry ("What AI powers Qaf?") to deflect this recurring question.
Threads
19fd4ce5eb524613— "OpenAI or Claude?" (question in subject line only) — answered: OpenAI19fdfff9d24dbea5— Which LLMs/providers, own models? — answered: agentset + GPT 5.6 sol, no own models