Language modelActive

Qwen 3 235B A22B

Alibaba

Released: April 27, 2025
Data date: September 8, 2026
Successor: Qwen 3 235B A22B Instruct 2507

Speed: 60 tokens/s
Output throughput
Price: $0.455 · $1.82
Input · Output / 1M tokens
Input: Text
Output: Text

The original Qwen 3 235B A22B checkpoint combines 235 billion total parameters with 22 billion active parameters and a switchable Thinking mode. Unlike the later separate instruct and thinking variants, it keeps both response modes in one checkpoint.

The architecture uses 128 experts, eight active per token, and 94 layers with 64 Q heads and 4 KV heads. Qwen lists 32,768 native context tokens and up to 131,072 tokens with YaRN. Unlike the separate Qwen 3 235B A22B Instruct 2507 and Qwen 3 235B A22B Thinking 2507, this original checkpoint combines both response modes.

Specifications and access

Specification	Value and source
Weights	Open weightsSource
Access	API, LocalSource
Parameters	235BSource
Active parameters	22BSource
Context window	32,768 tokensSource
Notes	The retained legacy slug refers to the original hybrid Qwen3-235B-A22B checkpoint. Its model card lists Pretraining and Post-training with switchable Thinking and non-Thinking modes. It is not a separately named Base checkpoint.Source
Weights repository	Qwen/Qwen3-235B-A22BSource
License	Apache-2.0Source
Architecture	Qwen3 causal language model; dense or sparse MoESource
Input	TextSource
Output	TextSource
Native context window	32,768 tokensSource
Maximum context window	131,072 tokensSource
Parameters	235 BSource
Active parameters	22 BSource
Languages	119 languages and dialectsSource
Repository created	2025-04-27 (Repository creation date is a precise publication proxy, but may differ from the public announcement date or an individual checkpoint's first availability.)Source
Training tokens	36 trillionSource

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

Artificial Analysis Intelligence Index (Show measurement, test conditions, and source)

Source value: 8.29469219577047
Score: 8.29469219577047
Metric: points
Unit: points
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. Non-reasoning

Source Artificial Analysis

AA-Omniscience Index (Show measurement, test conditions, and source)

Source value: -52.3833333333333
Score: -52.3833333333333
Metric: points
Unit: points
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. Non-reasoning

Source Artificial Analysis

Output speed (Show measurement, test conditions, and source)

Source value: 59.9542993186028
Score: 59.9542993186028
Metric: tokens/s
Unit: tokens/s
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. Non-reasoning; default 10K-input workload only
Context: Artificial Analysis performance timing; output speed is decoded after the first streaming chunk. Provider basis: first-party API. Deprecated model. Artificial Analysis only continues benchmarking the default 10K-input workload.

Source Artificial Analysis

Latency reasoning time (Show measurement, test conditions, and source)

Source value: 0
Score: 0
Metric: seconds
Unit: seconds
Direction: Lower is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. Non-reasoning; default 10K-input workload only
Context: Artificial Analysis performance timing; output speed is decoded after the first streaming chunk. Provider basis: first-party API. Deprecated model. Artificial Analysis only continues benchmarking the default 10K-input workload.

Source Artificial Analysis

Latency input time (Show measurement, test conditions, and source)

Source value: 2.647418257
Score: 2.647418257
Metric: seconds
Unit: seconds
Direction: Lower is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. Non-reasoning; default 10K-input workload only
Context: Artificial Analysis performance timing; output speed is decoded after the first streaming chunk. Provider basis: first-party API. Deprecated model. Artificial Analysis only continues benchmarking the default 10K-input workload.

Source Artificial Analysis

Answer generation time (end-to-end timing breakdown) (Show measurement, test conditions, and source)

Source value: 8.339685488491039
Score: 8.339685488491039
Metric: seconds
Unit: seconds
Direction: Lower is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. Non-reasoning; default 10K-input workload only
Context: Artificial Analysis performance timing; output speed is decoded after the first streaming chunk. Provider basis: first-party API. Deprecated model. Artificial Analysis only continues benchmarking the default 10K-input workload.

Source Artificial Analysis

End-to-end reasoning time component (Show measurement, test conditions, and source)

Source value: 0
Score: 0
Metric: seconds
Unit: seconds
Direction: Lower is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. Non-reasoning; default 10K-input workload only
Context: Artificial Analysis performance timing; output speed is decoded after the first streaming chunk. Provider basis: first-party API. Deprecated model. Artificial Analysis only continues benchmarking the default 10K-input workload.

Source Artificial Analysis

End-to-end input time component (Show measurement, test conditions, and source)

Source value: 2.647418257
Score: 2.647418257
Metric: seconds
Unit: seconds
Direction: Lower is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. Non-reasoning; default 10K-input workload only
Context: Artificial Analysis performance timing; output speed is decoded after the first streaming chunk. Provider basis: first-party API. Deprecated model. Artificial Analysis only continues benchmarking the default 10K-input workload.

Source Artificial Analysis

Predecessor and successor

Sources and data date

Type	Evidence and data date
Speed	Alibaba, Non-reasoning. Output speed after the first streaming chunk, not total response time. Classified as deprecated by Artificial Analysis. Artificial Analysis maintains only the default 10K-input workload there. Retrieved September 8, 2026. Source: Artificial Analysis · Artificial Analysis (retrieved September 8, 2026)Speed measurement methodology
Price	Base API prices without caching or batch discounts. Higher context tiers and other rates are listed under Costs. Source: Pricing · API pricing
Input	Documented modalities. Retrieved September 8, 2026. Source: Model documentation · Model metadata (retrieved September 8, 2026) · Qwen3-235B-A22B post-trained model card · Editorial description reviewed October 2, 2026
Output	Documented modalities. Retrieved September 8, 2026. Source: Model documentation · Model metadata (retrieved September 8, 2026) · Qwen3-235B-A22B post-trained model card · Editorial description reviewed October 2, 2026
Research status	Research date September 8, 2026. 4 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not found in the inspected sources	excluded-checkpoint-observations Original wording This profile describes the original post-trained Qwen3-235B-A22B. The report's separate Base results do not measure this checkpoint.
Additional source	Model metadata (retrieved September 8, 2026) · Qwen3 Technical Report · Editorial description reviewed October 2, 2026
Additional source	Model metadata (retrieved September 8, 2026)
Additional source	Qwen3-235B-A22B Instruct 2507 model card · Editorial description reviewed October 2, 2026
Additional source	Qwen3-235B-A22B Thinking 2507 model card · Editorial description reviewed October 2, 2026
Additional source	OpenRouter: Popularity index (derived), CC BY 4.0
Additional source	Vercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional source	Creative Commons Attribution 4.0

Qwen 3 235B A22B

Specifications and access

Measurements without matching peer values

Our tests

Pelican on a bicycle

Popularity

Published prices

Predecessor and successor

Sources and data date