AlibabaLanguage model
Qwen2.5-32B-Instruct
Qwen2.5-32B-Instruct has no overall index value yet because there are too few independent benchmark results.
PopularityRank 127 of 785
49out of 100-1 vs. last month
based on search, usage and attention
Summary
Qwen2.5-32B-Instruct is the general 32B configuration in the series. Its model card highlights stronger instruction following, long text generation, table understanding, and structured JSON output. It remains a text-only model and should not be confused with the specialized Coder line.
The repository supports 131,072 context tokens and up to 8,192 generated tokens. Apache 2.0 permits self-hosting. The parallel variant specialized for source code is Qwen 2.5 Coder 32B.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 88 published benchmark results
Page 1 of 8
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.
RULER (Show measurement, test conditions, and source)
- Source value
- 92.9
- Score
- 92.9
- Metric
- average accuracy
- Unit
- %
- Benchmark version
- Qwen3 Technical Report v1, Table 23
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Table 23; exact model column Qwen2.5-32B-Instruct; RULER. YARN scaling_factor=4 is used for length extrapolation. These are Qwen2.5 instruction checkpoints listed in a later Qwen3 report, not Qwen3 scores.
- Context
- The source does not report a scored sample count or confidence interval for this cell.
RULER 4K (Show measurement, test conditions, and source)
- Source value
- 96.9
- Score
- 96.9
- Metric
- accuracy at context length
- Unit
- %
- Benchmark version
- Qwen3 Technical Report v1, Table 23
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Table 23; exact model column Qwen2.5-32B-Instruct; RULER. YARN scaling_factor=4 is used for length extrapolation. These are Qwen2.5 instruction checkpoints listed in a later Qwen3 report, not Qwen3 scores.
- Context
- The source does not report a scored sample count or confidence interval for this cell.
RULER 8K (Show measurement, test conditions, and source)
- Source value
- 97.1
- Score
- 97.1
- Metric
- accuracy at context length
- Unit
- %
- Benchmark version
- Qwen3 Technical Report v1, Table 23
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Table 23; exact model column Qwen2.5-32B-Instruct; RULER. YARN scaling_factor=4 is used for length extrapolation. These are Qwen2.5 instruction checkpoints listed in a later Qwen3 report, not Qwen3 scores.
- Context
- The source does not report a scored sample count or confidence interval for this cell.
RULER 16K (Show measurement, test conditions, and source)
- Source value
- 95.5
- Score
- 95.5
- Metric
- accuracy at context length
- Unit
- %
- Benchmark version
- Qwen3 Technical Report v1, Table 23
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Table 23; exact model column Qwen2.5-32B-Instruct; RULER. YARN scaling_factor=4 is used for length extrapolation. These are Qwen2.5 instruction checkpoints listed in a later Qwen3 report, not Qwen3 scores.
- Context
- The source does not report a scored sample count or confidence interval for this cell.
RULER 32K (Show measurement, test conditions, and source)
- Source value
- 95.5
- Score
- 95.5
- Metric
- accuracy at context length
- Unit
- %
- Benchmark version
- Qwen3 Technical Report v1, Table 23
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Table 23; exact model column Qwen2.5-32B-Instruct; RULER. YARN scaling_factor=4 is used for length extrapolation. These are Qwen2.5 instruction checkpoints listed in a later Qwen3 report, not Qwen3 scores.
- Context
- The source does not report a scored sample count or confidence interval for this cell.
RULER 64K (Show measurement, test conditions, and source)
- Source value
- 90.3
- Score
- 90.3
- Metric
- accuracy at context length
- Unit
- %
- Benchmark version
- Qwen3 Technical Report v1, Table 23
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Table 23; exact model column Qwen2.5-32B-Instruct; RULER. YARN scaling_factor=4 is used for length extrapolation. These are Qwen2.5 instruction checkpoints listed in a later Qwen3 report, not Qwen3 scores.
- Context
- The source does not report a scored sample count or confidence interval for this cell.
RULER 128K (Show measurement, test conditions, and source)
- Source value
- 82
- Score
- 82
- Metric
- accuracy at context length
- Unit
- %
- Benchmark version
- Qwen3 Technical Report v1, Table 23
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Qwen3 Technical Report
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. Table 23; exact model column Qwen2.5-32B-Instruct; RULER. YARN scaling_factor=4 is used for length extrapolation. These are Qwen2.5 instruction checkpoints listed in a later Qwen3 report, not Qwen3 scores.
- Context
- The source does not report a scored sample count or confidence interval for this cell.
VerifierBench (binary, overall average) (Show measurement, test conditions, and source)
- Source value
- 63.8
- Score
- 63.8
- Metric
- Accuracy
- Unit
- %
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- CompassVerifier: A Unified and Robust Verifier for LLMs Evaluation and Outcome Reward (EMNLP 2025)
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. The paper names the exact instruction checkpoint Qwen2.5-32B-Instruct. Appendix A.3 says open-weight models use vLLM, evaluation temperature is 0, max_gen_len and max_model_len are set to their maximum values, and general LLMs use the authors' non-CoT prompt. The paper does not state an immutable repository revision or a serving route beyond vLLM.
- Context
- The source reports a 2,817-sample dataset, not a model-specific scored N; the paper provides no per-model N or confidence interval in the cited tables. It does not pin the evaluated weight revision.
VerifierBench (binary, overall average) (Show measurement, test conditions, and source)
- Source value
- 42.2
- Score
- 42.2
- Metric
- F1
- Unit
- %
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- CompassVerifier: A Unified and Robust Verifier for LLMs Evaluation and Outcome Reward (EMNLP 2025)
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. The paper names the exact instruction checkpoint Qwen2.5-32B-Instruct. Appendix A.3 says open-weight models use vLLM, evaluation temperature is 0, max_gen_len and max_model_len are set to their maximum values, and general LLMs use the authors' non-CoT prompt. The paper does not state an immutable repository revision or a serving route beyond vLLM.
- Context
- The source reports a 2,817-sample dataset, not a model-specific scored N; the paper provides no per-model N or confidence interval in the cited tables. It does not pin the evaluated weight revision.
VerifierBench (three-label, overall average) (Show measurement, test conditions, and source)
- Source value
- 58
- Score
- 58
- Metric
- Accuracy
- Unit
- %
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- CompassVerifier: A Unified and Robust Verifier for LLMs Evaluation and Outcome Reward (EMNLP 2025)
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. The paper names the exact instruction checkpoint Qwen2.5-32B-Instruct. Appendix A.3 says open-weight models use vLLM, evaluation temperature is 0, max_gen_len and max_model_len are set to their maximum values, and general LLMs use the authors' non-CoT prompt. The paper does not state an immutable repository revision or a serving route beyond vLLM.
- Context
- The source reports a 2,817-sample dataset, not a model-specific scored N; the paper provides no per-model N or confidence interval in the cited tables. It does not pin the evaluated weight revision.
VerifierBench (three-label, overall average) (Show measurement, test conditions, and source)
- Source value
- 45.4
- Score
- 45.4
- Metric
- Macro-F1
- Unit
- %
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- CompassVerifier: A Unified and Robust Verifier for LLMs Evaluation and Outcome Reward (EMNLP 2025)
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. The paper names the exact instruction checkpoint Qwen2.5-32B-Instruct. Appendix A.3 says open-weight models use vLLM, evaluation temperature is 0, max_gen_len and max_model_len are set to their maximum values, and general LLMs use the authors' non-CoT prompt. The paper does not state an immutable repository revision or a serving route beyond vLLM.
- Context
- The source reports a 2,817-sample dataset, not a model-specific scored N; the paper provides no per-model N or confidence interval in the cited tables. It does not pin the evaluated weight revision.
VerifyBench (Model-specific Prompt column) (Show measurement, test conditions, and source)
- Source value
- 78.8
- Score
- 78.8
- Metric
- Accuracy
- Unit
- %
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- CompassVerifier: A Unified and Robust Verifier for LLMs Evaluation and Outcome Reward (EMNLP 2025)
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. The paper's Table 2 defines the “Model-specific Prompt” column as the authors' own prompt for general LLMs (model-specific prompts apply to xVerify/Tencent RLVR), while the “VerifyBench Prompt” column uses the standard VerifyBench prompt for every model. The paper names the exact instruction checkpoint Qwen2.5-32B-Instruct. Appendix A.3 says open-weight models use vLLM, evaluation temperature is 0, max_gen_len and max_model_len are set to their maximum values, and general LLMs use the authors' non-CoT prompt. The paper does not state an immutable repository revision or a serving route beyond vLLM.
- Context
- The cited table gives no per-model N or confidence interval. “Model-specific Prompt” is the source column label; for general LLMs the paper says it uses the authors' prompt.
Page 1 of 8
Popularity compared
How often the model is searched for, used and mentioned, compared with every model of its category.
Popularity index
The 100 most popular models of this category, 0 to 100, higher is more popular
20 of 100 models
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 8, 2026
Qwen2.5-32B-InstructGLM-5.3 49Ling-3.0-tiny 49Qwen 3.5 122B A10B 49Qwen 3 14B 48
Method changes in the timeline, a jump at these points is no market event:
- 1November 1, 2026: Value only with search or usage data, 30-day attention
What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.
Similar models
More models of the same kind as Qwen2.5-32B-Instruct, chosen by model line, developer, size, price and release date.
- Qwen 2.5 Coder 32BAlibaba · same developer
- QwQ 32BAlibaba · same developer
- Qwen 3 30B A3BAlibaba · same developer
- Qwen 3 Coder 30B A3BAlibaba · same developer
- Qwen 3 VL 30B A3B ThinkingAlibaba · same developer
- Qwen 3 30B A3B Thinking 2507Alibaba · same developer
- Qwen 3 VL 30B A3BAlibaba · same developer
- Qwen 2.5 72BAlibaba · same developer
All specifications
Architecture
Input and output
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
Frequently asked questions about Qwen2.5-32B-Instruct
How large is the context window of Qwen2.5-32B-Instruct?
131,072 tokens. That is how much text Qwen2.5-32B-Instruct reads at once, about 200 A4 pages.
Page count as a rule of thumb from Artificial Analysis
GLM-5.1200K
DeepSeek-V3 0324164K
Qwen2.5-32B-InstructRank 94 of 103131K
GLM-4.5131K
GPT OSS 120B131K
DeepSeek-V3.2128K
DeepSeek-R1128K
Can I run Qwen2.5-32B-Instruct locally?
Yes, the weights are open.
Stated by Alibaba
- Open weights
- yes
- License
- Apache 2.0
- Parameters
- 32B
- active per token
- n/a
Sources and data date
Show all 24 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.
Without vendor claims and historical or disputed results; those stay labelled in the single values.