Playground

Test, compare, and choose your primary model — issues and all.

1. Showroom
2. Playground · pilot
3. Purchase & use · not yet available

Showroom · Before you buy

Qwen3-0.6B

Apache-2.0 · Hugging Face / Featherless AI

0.6B parameters · Arabic and multiple dialects among 119 languages.

Before you buy · Issues & all

Strengths: Small footprint; Arabic responses arrived quickly in Souq's pilot. A candidate for simple drafting tasks.
Known issues & limits: In an earlier pilot it gave the wrong Saudi capital and stray characters appeared. Fluency is not factual accuracy.
Dialect verdict: No human dialect audit yet; test your own dialect and tasks before choosing.

Test at your own risk · Verify outputs

Do not enter sensitive data or rely on it alone for medical, legal or financial decisions. This notice does not waive your legal rights.

Official source & licence

Qwen2.5-7B-Instruct

Apache-2.0 · Hugging Face / Featherless AI

7.61B parameters · Arabic among 29+ supported languages.

Before you buy · Issues & all

Strengths: Official Arabic support among 29+ languages. The developer reports improved instruction following and structured output.
Known issues & limits: Larger does not mean infallible: it can invent facts or mishandle dialects. Speed and cost depend on the provider.
Dialect verdict: No human dialect audit yet; test your own dialect and tasks before choosing.

Test at your own risk · Verify outputs

Do not enter sensitive data or rely on it alone for medical, legal or financial decisions. This notice does not waive your legal rights.

Official source & licence

Same prompt · Two models

My primary model

For registered members: 5 comparisons a day, 10 seconds between tests. Limit: 2,000 characters and 300 tokens per answer; replies may be truncated.

Qwen3-0.6B

No result yet

Qwen2.5-7B-Instruct

No result yet

Your primary-model choice is saved to your account; it does not grant an API key.

Purchase & use

Sales are not enabled yet: supplier terms, pricing and payments need confirmation — not a flawless model.

Before paying

Your chosen model, honest review and your test results.
Token pricing, spending cap, licence, data handling and cancellation terms.
Acknowledgement of limitations without waiving legal rights.

After paying · planned

Credit and your own Souq access key, never the Hugging Face token.
One OpenAI-compatible connection for your primary model.
Usage, bills and key revocation in your account.

Request paid access

Owner setup

Add Hugging Face credits and monitor spending; comparison uses two requests.
Create a fine-grained token with Inference Providers call permission.
Save HUGGINGFACE_API_KEY in project secrets, never chat. The token stays server-side.
Signed-in members can compare. Publish changes to update the live site. Provider pricing (opens in a new tab on huggingface.co)

Keep your comparisons

One-time purchases; no automatic renewal. These are downloads, not API keys. USD prices exclude tax shown at checkout.

I request digital access immediately after payment and understand the access period and Playground limits, without waiving my legal rights. Purchase & privacy terms

Playground download

Fixed archive of comparisons saved when checkout starts
Re-download the archive for 30 days after payment

Playground bundle (30 days)

Unlimited re-downloads for 30 days
Your best-model summary and each comparison's winner
20 comparisons a day instead of 5, for 30 days

This week in Arabic AI

New tools and the top news, once a week.