Qwen 3 32B
Reasoning tokens were reported. The output is not shown as a comparable short-test result.
Language modelActive
Qwen 3 32B
Alibaba
- Released
- April 27, 2025
- Data date
- September 8, 2026
- Predecessor
- Qwen 2.5 72B
- Speed
- 105.2 tokens/sOutput throughput
- Price
- Input · Output / 1M tokens
- Input
- Text
- Output
- Text
Qwen 3 32B brings the Qwen3 family into a dense 32.8-billion-parameter model with switchable Thinking. It can produce additional reasoning steps or answer directly without that visible trace.
The model has 64 layers with grouped-query attention, 64 Q heads, and 8 KV heads. Qwen lists 32,768 native context tokens and up to 131,072 tokens with YaRN.
Specifications and access
| Specification | Value and source |
|---|---|
| Weights | Open weightsSource |
| Access | API, LocalSource |
| Parameters | 32.8BSource |
| Context window | 32,768 tokensSource |
| Weights repository | Qwen/Qwen3-32BSource |
| License | Apache-2.0Source |
| Architecture | Qwen3 causal language model; dense or sparse MoESource |
| Input | TextSource |
| Output | TextSource |
| Reasoning mode | base checkpoint; report evaluates pre-training onlySource |
| Tool calling | NoSource |
| Structured output | NoSource |
| Native context window | 32,768 tokensSource |
| Maximum context window | 131,072 tokensSource |
| Parameters | 32.8 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 |
Page 1 of 2
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.
MMLU (Show measurement, test conditions, and source)
- Source value
- 83.61
- Score
- 83.61
- Metric
- percent
- Unit
- %
- Category
- knowledge
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
MMLU-Redux (Show measurement, test conditions, and source)
- Source value
- 83.41
- Score
- 83.41
- Metric
- percent
- Unit
- %
- Category
- knowledge
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
MMLU-Pro (Show measurement, test conditions, and source)
- Source value
- 65.54
- Score
- 65.54
- Metric
- percent
- Unit
- %
- Category
- knowledge
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
SuperGPQA (Show measurement, test conditions, and source)
- Source value
- 39.78
- Score
- 39.78
- Metric
- percent
- Unit
- %
- Category
- knowledge
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
BBH (Show measurement, test conditions, and source)
- Source value
- 87.38
- Score
- 87.38
- Metric
- percent
- Unit
- %
- Category
- knowledge
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
GPQA (Show measurement, test conditions, and source)
- Source value
- 49.49
- Score
- 49.49
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
GSM8K (Show measurement, test conditions, and source)
- Source value
- 93.4
- Score
- 93.4
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
MATH (Show measurement, test conditions, and source)
- Source value
- 61.62
- Score
- 61.62
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
EvalPlus (Show measurement, test conditions, and source)
- Source value
- 72.05
- Score
- 72.05
- Metric
- percent
- Unit
- %
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
MultiPL-E (Show measurement, test conditions, and source)
- Source value
- 67.06
- Score
- 67.06
- Metric
- percent
- Unit
- %
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
MBPP (Show measurement, test conditions, and source)
- Source value
- 78.2
- Score
- 78.2
- Metric
- percent
- Unit
- %
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
CRUX-O (Show measurement, test conditions, and source)
- Source value
- 72.5
- Score
- 72.5
- Metric
- percent
- Unit
- %
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official Qwen3 base-model evaluation; benchmark protocol in report table notes
Page 1 of 2
Our tests
SVG short test
One SVG prompt and fixed output settings. Captures are shown only with 0 reported reasoning tokens or documented lack of reasoning controls. Identical test conditions do not make the models' internal compute identical.
Earlier tests
Pelican on a bicycle
A community classic for free-form SVG drawing.
Qwen 3 32B
Reasoning (requested): high
Current attempt
- Run date
- Oct 1, 2026
- API provider
- openrouter
- Requested API model
- qwen/qwen3-32b
- API model reported in response
- qwen/qwen3-32b
- Reasoning setting
- high
- Pinned endpoint
- DeepInfra | qwen/qwen3-32b-04-28
- Temperature
- Provider default
- Token limit including reasoning
- 8,192 tokens
- Test protocol
- community-visual-tests-v1
- Technical checks
- Passed, not a quality rating
- Response time
- 52.8 s
Visually reviewed. This individual test does not establish overall model quality.
Prompt and test conditions
Generate an SVG of a pelican riding a bicycle
Token limit including reasoning: 8,192 tokens
Task origin (Simon Willison)Popularity
44.3 / 100
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. |
| Price | Base API prices without caching or batch discounts. Higher context tiers and other rates are listed under Costs. |
| Input | Documented modalities. Retrieved September 8, 2026. |
| Output | Documented modalities. Retrieved September 8, 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 | active-parameters-billions Original wording Dense checkpoint; no separate activated-parameter count is applicable/published. |
| Additional source | Qwen3 Technical Report (retrieved September 8, 2026) · Qwen3 Technical Report · Editorial description reviewed October 2, 2026 |
| Additional source | Model metadata (retrieved September 8, 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 |
| Additional source | Qwen: "Qwen3: Think Deeper, Act Faster" (2026-09-19) |