Qwen3 235B-A22B — Spécifications
| Développeur | Alibaba |
|---|---|
| Type | LLM (architecture MoE) |
| Modalité | Texte → Texte |
| Paramètres | 235 milliards au total / 22 milliards actifs (MoE) |
| Fenêtre de contexte | 128 K |
| Sortie maximale | — |
| Licence | Apache 2.0 (ouverte) |
| Poids ouverts | Oui |
| Publié | 2025 |
| Prix de l’entrée | 0,45 $ / 1 million |
| Prix de la sortie | 1,80 $ / 1 million |
| Fournisseurs d'API | Alibaba, OpenRouter |
Exécutez-le localement
| VRAM (4 bits) | ~140 Go |
|---|---|
| GPU minimal requis | Multi-GPU ou Mac avec 192 Go |
What is Qwen3 235B-A22B?
Qwen3 235B-A22B is the open flagship of Alibaba’s Qwen3 family — a 235B mixture-of-experts
activating 22B parameters per token, released under Apache 2.0 with a 128K context. Pricing
is $0.45 in / $1.80 out per million tokens; self-hosting needs around 140 GB at 4-bit, which
means a multi-GPU setup or a 192 GB Mac.
The Mac line is not a curiosity. Apple’s unified memory architecture lets a single
workstation address far more model memory than any consumer GPU, so a 192 GB Mac Studio can
hold a 235B model that would otherwise require a multi-GPU server — at a fraction of the
power draw, noise and facilities cost. For a small team that needs frontier-adjacent open
weights on-premises, that is often the only realistic path, and Qwen3 235B-A22B is one of the
strongest models that fits it. Apache 2.0 licensing makes it cleaner to adopt than the Llama 4
models, which carry EU restrictions. The 128K context is narrower than the 1M offered by
newer flagships, which is the main trade-off to weigh.
Qwen3 235B-A22B pricing: API cost per 1M tokens
| Entrée (par million de jetons) | $0.450 |
|---|---|
| Sortie (par million de jetons) | $1.80 |
| Output/input ratio | 4× |
| Blended (4:1 in:out) | $0.720 per 1M tokens |
What Qwen3 235B-A22B costs per month
Real monthly spend at a 4:1 input-to-output mix — the ratio a typical chat or RAG workload actually produces.
| Charge de travail | Jetons/mois | Cost / month |
|---|---|---|
| Projet secondaire | 1 million en entrée / 0,25 million en sortie | $0.90 |
| Petite équipe | 20 millions en entrée / 5 millions en sortie | $18 |
| Production | 200 millions en entrée / 50 millions en sortie | $180 |
Run your own numbers in the Calculateur de coûts des API IA.
Cheaper alternatives to Qwen3 235B-A22B
| Modèle | Coût combiné par million de dollars | You save |
|---|---|---|
| DeepSeek V4-Pro ouverte | $0.522 | 28% cheaper |
| DeepSeek V4-Flash ouverte | $0.168 | 77% cheaper |
| Llama 4 Maverick ouverte | $0.240 | 67 % moins cher |
Self-host or pay the API?
Qwen3 235B-A22B is open-weight, so you can run it yourself. It needs ~140 Go of VRAM at 4-bit (Multi-GPU or Mac 192GB). Self-hosting only beats the API once your volume is high enough to keep that hardware busy — the calculateur auto-hébergement vs API works out the break-even point for your token volume.
Questions fréquemment posées
How much does Qwen3 235B-A22B cost per 1M tokens?
Qwen3 235B-A22B costs $0.450 per 1M input tokens and $1.80 per 1M output tokens. At a typical 4:1 input-to-output mix that blends to about $0.720 per 1M tokens.
How much does Qwen3 235B-A22B cost per month?
A small-team workload of 20M input and 5M output tokens a month costs about $18 on Qwen3 235B-A22B. A side project (1M in / 0.25M out) costs roughly $0.90.
What is a cheaper alternative to Qwen3 235B-A22B?
DeepSeek V4-Pro is the strongest cheaper option in our database at $0.522 per 1M blended — about 28% less than Qwen3 235B-A22B. It is also open-weight, so self-hosting is an option.
Can I run Qwen3 235B-A22B locally?
Yes. Qwen3 235B-A22B is open-weight and needs about ~140 GB of VRAM at 4-bit quantisation (Multi-GPU or Mac 192GB).
Why does Qwen3 235B-A22B charge more for output than input?
Output tokens are generated one at a time and cannot be batched the way a prompt can, so they cost the provider more to serve. Qwen3 235B-A22B charges 4× more for output, which is why prompt-heavy workloads are far cheaper to run than generation-heavy ones.
Prices are the published list rates for the model's primary API and are reviewed as providers change them. Volume, batch and cached-input discounts are not included. Compare every model side by side in the Base de données des modèles d'IA ou le Classement des grands modèles linguistiques (LLM).
