See how ChatGPT, Perplexity and Google AI Overviews describe you today

Free AI Visibility Audit
Surge45°
Qwen model°

Qwen3 235B A22B

Qwen3 235B A22B is Qwen's large open-weight mixture-of-experts text model, which activates a fraction of its total parameters on each pass and can switch between a "thinking" mode for harder reasoning and a faster direct mode.

QwenVerified 24/09/2026Released 28/04/2025

01 / decision

Decision snapshot

Each figure sits next to the middle of the field, so it can be read as dear or cheap, wide or narrow, rather than floating on its own. Compared against the 126 models we track, not against an absolute standard.

Capability

Not measured

No published benchmark scores yet

Input / 1M tokens

$0.46

field median $0.30

Output / 1M tokens

$1.82

field median $1.25

Context

131K

field median 524K

02 / overview

What Qwen3 235B A22B is, and when to reach for it

Four questions, answered separately, because somebody arrives at one of them rather than at the top of the page.

What it is

The job it was built for, and the job it is not for.

Qwen3 235B A22B is a text-only reasoning and general-purpose model built around a mixture-of-experts design, so a large parameter count sits behind a much smaller active compute cost per request. The notable feature is the switch between thinking mode, aimed at complex reasoning and maths, and ordinary response mode for everything else. It handles no images, audio or video, so anything involving documents as pictures or screenshots is out of scope.

When it arrived, and when to use it

Where it sits in its line, and when a sibling is the better pick.

Qwen3 235B A22B first appeared in late April 2025 as the flagship of the Qwen3 line, sitting above the smaller dense Qwen3 models. It is the right pick when you want a single model that can be told to reason hard on the queries that need it and answer plainly on the ones that do not, without swapping endpoints. It is the wrong pick if your workload is multimodal, or if you need very long generated outputs in one call, since the output ceiling is modest relative to how much you can feed in.

How you reach it

The API, the apps it powers, and what its limits let you do.

Qwen3 235B A22B is reached through the Qwen API and through the several hosting providers that serve the open weights, and it can be self-hosted if you have the hardware. The large input window means you can put long documents, extended chat history or a sizeable code context in front of it in one go, while the tighter output limit points it towards answers, summaries and structured extraction rather than long-form generation. The thinking mode is a request-level choice, so routing logic in your own application decides which queries get the slower, more deliberate path.

Why it matters

What changes because this exists, or why it does not.

The mixture-of-experts structure is the point: teams get a frontier-sized model while paying compute for only the active portion, which makes sustained reasoning workloads affordable rather than a special occasion. Being able to toggle reasoning per request also removes the usual awkwardness of maintaining two models, one cheap and one careful, and deciding at the gateway which one a user's question deserves.

Follows Qwen3 14B. Superseded by Qwen3 30B A3B. See the whole line.

03 / evidence

How much of this is verified

Split by category, so a strong number never hides a thin evidence base. Verified means we read it on the benchmark's own published results; a provider's claim about its own model is shown and labelled rather than dropped.

No published benchmark scores for this model yet.

We publish a score only where we can link the result to where it was published. Until a benchmark result for this model exists in a source we read, this section stays empty rather than being filled with a provider's marketing figure.

How we decide what counts as evidence

04 / ledger

Benchmark ledger

Every published row, grouped by category, each compared with the best published score on the same benchmark. 'Is 64% good' is a question nobody can answer; '26 points behind the leader' is one anybody can.

Nothing in the ledger yet.

Each row here carries a score, the benchmark it came from, what the leading model scored on the same test, and a link to the published result. Rows appear as results are published and read.

How we decide what counts as evidence

05 / capability

Capability shape

Where this model is strong, and against how many peers. Ranks are against models with evidence in that category, not against everything we track: ranking against models nobody tested would rank who published, not who is better.

No category scores to shape yet.

A category score is the weighted mean of the benchmarks published for it. With no published rows there is nothing to average, and an empty chart drawn at zero would say something false.

How we decide what counts as evidence

06 / cost

What it costs

List API rates as last read from the provider, with the source on every row, plus every change we have recorded since we started tracking it.

Qwen3 235B A22B API pricingSurge45°
ChargePriceUnitRead onSource
Input$0.46per 1M tokens2026-09-24Check
Output$1.82per 1M tokens2026-09-24Check

Qwen3 235B A22B sits in the low-cost tier, well under what closed frontier reasoning models charge, with output priced at four times input in the usual pattern. For high-volume work where you want reasoning switched on by default, the cost profile is the main reason to consider it over a proprietary alternative.

What a month costsSurge45°
WorkloadInput / monthOutput / monthCost
A small product team20M tokens5M tokens$18.20
A busy support assistant200M tokens40M tokens$163.80
A document pipeline1000M tokens100M tokens$637.00

List API rates, no caching and no batch discount, which both providers offer and which change the answer a great deal. Treat these as the ceiling, not the bill.

07 / specs

Specifications

As listed by the provider's own catalogue and re-read every few hours. Anything absent is absent there too.

SpecificationSurge45°
Context window131,072 tokens
Maximum output8,192 tokens
Modalitiestext
Released28/04/2025
StatusCurrent
Catalogue identifierqwen/qwen3-235b-a22b

08 / lineage

Lineage

What this model replaced, what replaced it, and what else its provider has in the field.

Also from Qwen

09 / line

The line

Every model in this family in release order, so a page from eight months ago says in one glance that two newer ones exist.

Came before

← Qwen3 14B

Came after

Qwen3 30B A3B →
  1. 01Qwen3 14B28/04/2025
  2. 02Qwen3 235B A22B28/04/2025
  3. 03Qwen3 30B A3B28/04/2025
  4. 04Qwen3 32B28/04/2025
  5. 05Qwen3 8B28/04/2025
  6. 06Qwen3.7 Flash27/07/2026
  7. 07Qwen3.8 2.4T A95B12/08/2026
  8. 08Qwen3.8 27B14/08/2026
  9. 09Qwen3.8 Flash26/08/2026
  10. 10Qwen3.8 Max (0902)03/09/2026
  11. 11Qwen3.8 Omni Flash21/09/2026
  12. 12Qwen3.8 Max Prime23/09/2026

Ordered by release date and worked out from the naming, so a new member slots in as soon as its page exists. A retired model keeps its page and its place in the line.

10 / notes

Our notes

What this model changes for a brand trying to be cited in AI answers, and every change we have logged since it launched.

What it changes for you

Cheap open-weight reasoning models mean more products can afford to think before they answer, so more of the AI surfaces citing your brand will be doing multi-step retrieval and comparison rather than one-shot recall. That rewards content that holds up under scrutiny: clear claims, stated specifics, pages that answer the sub-questions a reasoning pass will generate. It also means you cannot know which model is reading you, so writing for one vendor's quirks is a poor use of time.

Where buyers meet this model

Buyers meet Qwen3 235B A22B through the Qwen API and through third-party inference platforms that host the open weights, rather than through a single branded consumer app in Western markets. More often it reaches them indirectly, as the model behind someone else's chat product, internal assistant or AI search feature, where the vendor chose it for the cost of running reasoning at scale.

Change log

Nothing published here yet. Changes appear within hours of a provider announcing them.

11 / questions

Questions

The things people ask about this model, answered from what is on this page rather than from anywhere else.

Can Qwen3 235B A22B read images or PDFs?
No. It is a text-only model, so anything arriving as an image, scan or screenshot needs a separate extraction step before it reaches the model.
What does the thinking mode actually change?
It switches the model into a slower, more deliberate path intended for complex reasoning and maths. You choose it per request, so straightforward queries can still take the fast route.
Is it suitable for long-form writing?
Less so. The input window is large but the maximum output is comparatively short, which suits answers, analysis and extraction better than generating long documents in a single call.
Surge45°

Is Qwen3 235B A22B recommending you?

Models change what gets cited. We measure whether AI answers name your brand or your competitors across every assistant, and show you what to change.

Free AI visibility audit Book a call