opendunes

Qwen3.8 2.4T A95B

by QwenReleased 12 August 2026

#10 for agent workThe open-weight variant of Qwen3.8 Max, a sparse mixture-of-experts with 95B active out of 2.4 trillion, suited to coding, research, complex reasoning and agent workflows, with a million-token context

TakesText
GivesText
Open in chatCreate API keyCompare

What it costs

You pay for what each request uses. No subscription, no minimum.

≈ 1,66 DA per message3 000 DA buys about 1 800 messages

20 % offVenice has cut its price. Every figure here is the new one.

One message, the ordinary use

1 000 tokens sent× 665,67 DA per 1M0,67 DA
500 tokens back× 1 997,01 DA per 1M1,00 DA
Total≈ 1,66 DA

Every rate

Meter↑ you send↓ comes back
Text tokens665,67 DA1 997,01 DA
Cached tokens read83,21 DA

per 1M tokens

Answer a quick question≈ 0,60 DA
Reply to a customer email≈ 1,20 DA
Translate a full page of text≈ 2,06 DA
Review a pull request of about 400 lines≈ 6,39 DA
Summarise a 20-page report≈ 10,72 DA
When you pay
After the answer, from your balance. Nothing is held.
Why it varies
A longer question or a longer answer costs more.
After each one
The exact charge shows on the request.
DepositDeposits from 50 DA, with Edahabia or CIB.

Where your request runs

Four companies run this model. OpenDunes sends each request to one of them, picked by price and speed. If it fails, the next one takes it. DA per 1M tokens: ↑ what you send · ↓ what comes back.

Novita…
↑ Sent, per 1M
665,67 DA
↓ Back, per 1M
1 997,01 DA
Cache read
83,21 DA
Latency
…
Throughput
…
AlibabaAsia…
↑ Sent, per 1M
665,67 DA
↓ Back, per 1M
1 997,01 DA
Cache read
83,21 DA
Latency
…
Throughput
…
SiliconFlowUnited States…
↑ Sent, per 1M
665,67 DA
↓ Back, per 1M
1 997,01 DA
Cache read
83,21 DA
Latency
…
Throughput
…
Venice20 % offUnited States…
↑ Sent, per 1M
Was 832,09665,67 DA
↓ Back, per 1M
Was 2 496,261 997,01 DA
Cache read
Was 104,0183,21 DA
Latency
…
Throughput
…
Provider↑ Sent↓ BackCache readLatencyThroughputUptime
Novita665,67 DA1 997,01 DA83,21 DA………
AlibabaAsia665,67 DA1 997,01 DA83,21 DA………
SiliconFlowUnited States665,67 DA1 997,01 DA83,21 DA………
Venice20 % offUnited StatesWas 832,09665,67 DAWas 2 496,261 997,01 DAWas 104,0183,21 DA………

Also charged: writing to the cache, at Alibaba ↑ 832,09.

You pay the price of the provider that served your request. Prices read 15:21 · speed over the last 30 minutes · uptime over 3 days.

See it in action

Real requests, real replies. Try any of them yourself — one click, pre-filled.

Structured output
You send

Turn this customer email into JSON with intent, urgency and order_id.

customer-email.txt
It replied

{ "intent": "refund", "urgency": "high", "order_id": "OD-58421" }

Try in Chat Lab
Writing
You send

Two lines to welcome a shop in Oran.

It replied

Welcome to our new shop in Oran — fresh bread every morning, and coffee the way the city likes it. Come in, the first cup is on us.

Try in Chat Lab

Try your own request — same model, pre-selected.

Open Chat Lab

Call it in code

Swap the base URL and the key. Your OpenAI code runs unchanged.

POST /v1/chat/completions
Base URL
https://opendunes.com/v1
Model ID
qwen/qwen3.8-2.4t-a95b
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://opendunes.com/v1",
    api_key=os.environ["OPENDUNES_API_KEY"],
)

resp = client.chat.completions.create(
    model="qwen/qwen3.8-2.4t-a95b",
    messages=[{"role": "user", "content": "Two lines to welcome a shop in Oran."}],
)

print(resp.choices[0].message.content)
pip install openaiFull reference for this route
Sign in to run it

Runs this request once on your balance.

Parameters you can send

Standard OpenAI fields, grouped by what they control. Anything not listed is accepted and ignored.

Shaping the reply

How long, how varied, where to stop.

max_tokenstemperaturetop_ptop_kstopseedfrequency_penaltypresence_penaltyrepetition_penaltylogprobstop_logprobs

Reasoning

Let it think before answering — and see the thinking.

reasoningreasoning_effortinclude_reasoning

Structured output

Get JSON you can parse instead of prose.

response_formatstructured_outputs

Tools

Let the model call your functions.

toolstool_choice

Questions about Qwen3.8 2.4T A95B

  • How much does Qwen3.8 2.4T A95B cost in dinars?

    About 1,66 DA per message. Was 2,08 DA. ↑ 665,67 DA · ↓ 1 997,01 DA per 1M tokens.

  • Is Qwen3.8 2.4T A95B free? Can I try it?

    No. You pay per use from your balance, with no subscription. Try it in Chat at the same price.

  • What is Qwen3.8 2.4T A95B best for?

    #10 for agent work · The open-weight variant of Qwen3.8 Max, a sparse mixture-of-experts with 95B active out of 2.4 trillion, suited to coding, research, complex reasoning and agent workflows, with a million-token context.

  • How do I pay for Qwen3.8 2.4T A95B without an international card?

    Deposit dinars with your Edahabia or CIB card. Every model, Qwen3.8 2.4T A95B included, is paid from that balance. No foreign card, no conversion.

  • Does Qwen3.8 2.4T A95B work in Arabic, French or Darja?

    Qwen names no languages for Qwen3.8 2.4T A95B. Arabic, French and Darja: not stated.

  • Can it read PDFs, images or a whole book?

    It reads text. Up to 1M tokens in one request, about 1 000 pages. It writes up to 98 000 words at once.

  • Is Qwen3.8 2.4T A95B open source?

    Yes, its weights are public. Open weights on Hugging Face

  • Who makes Qwen3.8 2.4T A95B, and when was it released?

    Qwen. Released 12 August 2026.

Other Qwen models, priced in dinars

See all Qwen models
Qwen3 Embedding 8BQwen · Oct 28, 2025

The 8B Qwen3 Embedding, carrying Qwen3's multilingual reach and long-text understanding into vectors for text and code retrieval, classification, clustering and bitext mining; the 4B is its smaller sibling

40 pages per call
≈0,0033 DAper page
View model
Qwen3 Reranker 8BQwen · Aug 13, 2026

Qwen3 Reranker 8B reads a search question together with each candidate text and gives each pair a relevance score, so the best matches rise to the top. It covers over 100 languages and programming languages, and you can tell it what counts as relevant for your task. Use it as the second pass after a search step such as Qwen3's text-meaning models has fetched candidates, to put your answers in order. It does not write text itself.

40 pages per search
≈0,067 DAper search
View model
Qwen3 Embedding 4BQwen · Oct 28, 2025

Qwen3 Embedding 4B turns text into a fingerprint of its meaning, so search finds text that means the same even in other words or another language. It covers over 100 languages, code included, and reads about 32 pages at once, which suits searching documents or code, grouping similar texts and matching translations.

32 pages per call
≈0,0067 DAper page
View model
Qwen-Audio-3.0-TTS PlusQwen · Jul 23, 2026

Qwen-Audio-3.0-TTS Plus reads text and speaks it aloud, and Alibaba's docs say it takes plain-language directions for speech rate, emotion and speaking style. It is the higher-quality tier of the two Qwen-Audio-3.0 speech models, made by Alibaba. Pick it for narration, voiceover or any line where voice quality matters more than speed.

2 voices
≈3,99 DAper paragraph read aloud
View model