DeepSeekLanguage model
DeepSeek-R1-Distill-Qwen-32B
DeepSeek-R1-Distill-Qwen-32B has no overall index value yet because there are too few independent benchmark results.
PopularityRank 80 of 785
52out of 100+12 vs. last month
based on search, usage and attention
Summary
DeepSeek-R1-Distill-Qwen-32B brings reasoning data generated by DeepSeek-R1 into a dense Qwen2.5-32B checkpoint. Its model card lists a 131,072-token context window and an MIT license; the R1 orientation comes without the original’s mixture-of-experts structure and scale.
The model card documents no native tool-calling interface.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 63 published benchmark results
Page 1 of 6
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.
IFBench: Overall (Show measurement, test conditions, and source)
- Source value
- 22.9251700680272
- Score
- 22.9251700680272
- Metric
- IFBench prompt-level loose accuracy (pass@1)
- Unit
- %
- Benchmark version
- AA test set: 294 questions
- Category
- general
- Task
- overall
- Task label
- Overall
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Artificial Analysis
- Status
- superseded
- Retrieved at
- 2026-10-19
- Methodology
- Exact AA configuration: DeepSeek R1 Distill Qwen 32B (release deepseek-r1-distill-qwen-32b, row deepseek-r1-distill-qwen-32b, UUID df95f83f-5ebb-466a-9d2d-b95efc8c012c, reasoning=true, source reasoning-token configuration budget=2000, suffix=not specified). This configured value is not measured token consumption. Artificial Analysis runs 294 questions from the IFBench test set with five repeats per question. The official evaluator scores prompt-level accuracy in loose mode as pass@1.. Source model ID: artificial-analysis/df95f83f-5ebb-466a-9d2d-b95efc8c012c/deepseek-r1-distill-qwen-32b/deepseek-r1-distill-qwen-32b. Source model name: DeepSeek R1 Distill Qwen 32B. Provider reported by source: DeepSeek. Model snapshot: DeepSeek R1 Distill Qwen 32B. Evaluator: Artificial Analysis. Dataset type: artificial-analysis-evaluation. Tool policy: Single-turn answer evaluated with the official IFBench source code. Attempts: pass@1, 5 repeats per question. Reasoning: reasoning enabled. Score display: 22.9%. Source ID: artificial-analysis. Evidence capture ID: aa-model-page-2026-10-19-8a022369738c. Source payload hash: 8a022369738cbe2a0e2f9bf486a2882865dbb0c6f1396dfec84059a802560e3c. Parser version: artificial-analysis-rsc-models-v2. Imported at: 2026-10-19. Source cohort: artificial-analysis-ifbench:AA test set: 294 questions:standalone:reasoning:tokens=2000:suffix=unspecified
- Comparison methodology
- Exact AA configuration: DeepSeek R1 Distill Qwen 32B (release deepseek-r1-distill-qwen-32b, row deepseek-r1-distill-qwen-32b, UUID df95f83f-5ebb-466a-9d2d-b95efc8c012c, reasoning=true, source reasoning-token configuration budget=2000, suffix=not specified). This configured value is not measured token consumption. Artificial Analysis runs 294 questions from the IFBench test set with five repeats per question. The official evaluator scores prompt-level accuracy in loose mode as pass@1.. Tool policy: Single-turn answer evaluated with the official IFBench source code. Attempts: pass@1, 5 repeats per question. Reasoning: reasoning enabled
- Context
- The often cited 58 refers to verifiable out-of-domain constraint types, not the number of questions in AA's test run.. This result is retained as historical evidence and is not interchangeable with current results.
IFBench: Overall (Show measurement, test conditions, and source)
- Source value
- 22.9251700680272
- Score
- 22.9251700680272
- Metric
- IFBench prompt-level loose accuracy (pass@1)
- Unit
- %
- Benchmark version
- AA test set: 294 questions
- Category
- general
- Task
- overall
- Task label
- Overall
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Artificial Analysis
- Status
- superseded
- Retrieved at
- 2026-11-02
- Methodology
- Exact AA configuration: DeepSeek R1 Distill Qwen 32B (release deepseek-r1-distill-qwen-32b, row deepseek-r1-distill-qwen-32b, UUID df95f83f-5ebb-466a-9d2d-b95efc8c012c, reasoning=true, effort=not specified, suffix=not specified). AA source value for average reasoning tokens before the answer: 2000. Its methodology averages 60 varied prompts and assumes 2,000 when no average is available. This is not a token cap. Artificial Analysis runs 294 questions from the IFBench test set with five repeats per question. The official evaluator scores prompt-level accuracy in loose mode as pass@1.. Source model ID: artificial-analysis/df95f83f-5ebb-466a-9d2d-b95efc8c012c/deepseek-r1-distill-qwen-32b/deepseek-r1-distill-qwen-32b. Source model name: DeepSeek R1 Distill Qwen 32B. Provider reported by source: DeepSeek. Model snapshot: DeepSeek R1 Distill Qwen 32B. Evaluator: Artificial Analysis. Dataset type: artificial-analysis-evaluation. Tool policy: Single-turn answer evaluated with the official IFBench source code. Attempts: pass@1, 5 repeats per question. Reasoning: reasoning enabled. Score display: 22.9%. Source ID: artificial-analysis. Evidence capture ID: aa-model-page-2026-11-02-67d46be20a5b. Source payload hash: 67d46be20a5b412db6cfb60ad190441fcaa5513500b8fd47d033168d0e97b6b2. Parser version: artificial-analysis-rsc-models-v3. Imported at: 2026-11-02. Source cohort: artificial-analysis-ifbench:AA test set: 294 questions:standalone:reasoning:effort=unspecified:suffix=unspecified
- Comparison methodology
- Exact AA configuration: DeepSeek R1 Distill Qwen 32B (release deepseek-r1-distill-qwen-32b, row deepseek-r1-distill-qwen-32b, UUID df95f83f-5ebb-466a-9d2d-b95efc8c012c, reasoning=true, effort=not specified, suffix=not specified). AA source value for average reasoning tokens before the answer: 2000. Its methodology averages 60 varied prompts and assumes 2,000 when no average is available. This is not a token cap. Artificial Analysis runs 294 questions from the IFBench test set with five repeats per question. The official evaluator scores prompt-level accuracy in loose mode as pass@1.. Tool policy: Single-turn answer evaluated with the official IFBench source code. Attempts: pass@1, 5 repeats per question. Reasoning: reasoning enabled
- Context
- The often cited 58 refers to verifiable out-of-domain constraint types, not the number of questions in AA's test run.. This result is retained as historical evidence and is not interchangeable with current results.
IFBench: Overall (Show measurement, test conditions, and source)
- Source value
- 22.9251700680272
- Score
- 22.9251700680272
- Metric
- IFBench prompt-level loose accuracy (pass@1)
- Unit
- %
- Benchmark version
- AA test set: 294 questions
- Category
- general
- Task
- overall
- Task label
- Overall
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Artificial Analysis
- Status
- active
- Retrieved at
- 2026-11-07
- Methodology
- Exact AA configuration: DeepSeek R1 Distill Qwen 32B (release deepseek-r1-distill-qwen-32b, row deepseek-r1-distill-qwen-32b, UUID df95f83f-5ebb-466a-9d2d-b95efc8c012c, reasoning=true, effort=not specified, suffix=not specified). This source capture does not provide an average reasoning-token count. Artificial Analysis runs 294 questions from the IFBench test set with five repeats per question. The official evaluator scores prompt-level accuracy in loose mode as pass@1.. Source model ID: artificial-analysis/df95f83f-5ebb-466a-9d2d-b95efc8c012c/deepseek-r1-distill-qwen-32b/deepseek-r1-distill-qwen-32b. Source model name: DeepSeek R1 Distill Qwen 32B. Provider reported by source: DeepSeek. Model snapshot: DeepSeek R1 Distill Qwen 32B. Evaluator: Artificial Analysis. Dataset type: artificial-analysis-evaluation. Tool policy: Single-turn answer evaluated with the official IFBench source code. Attempts: pass@1, 5 repeats per question. Reasoning: reasoning enabled. Score display: 22.9%. Source ID: artificial-analysis. Evidence capture ID: aa-model-page-2026-11-07-b96e3ffc8c5b. Source payload hash: b96e3ffc8c5b052aa56f96d05e89d02b8cc234a75085820fca0ca2a39e498274. Parser version: artificial-analysis-rsc-models-v4. Imported at: 2026-11-07. Source cohort: artificial-analysis-ifbench:AA test set: 294 questions:standalone:reasoning:effort=unspecified:suffix=unspecified
- Comparison methodology
- Exact AA configuration: DeepSeek R1 Distill Qwen 32B (release deepseek-r1-distill-qwen-32b, row deepseek-r1-distill-qwen-32b, UUID df95f83f-5ebb-466a-9d2d-b95efc8c012c, reasoning=true, effort=not specified, suffix=not specified). This source capture does not provide an average reasoning-token count. Artificial Analysis runs 294 questions from the IFBench test set with five repeats per question. The official evaluator scores prompt-level accuracy in loose mode as pass@1.. Tool policy: Single-turn answer evaluated with the official IFBench source code. Attempts: pass@1, 5 repeats per question. Reasoning: reasoning enabled
- Context
- The often cited 58 refers to verifiable out-of-domain constraint types, not the number of questions in AA's test run.
DocPuzzle overall accuracy (Show measurement, test conditions, and source)
- Source value
- 39.7
- Score
- 39.7
- Metric
- Process-aware long-context QA accuracy
- Unit
- %
- Benchmark version
- DocPuzzle, arXiv:2502.17807v1
- Category
- slow-thinking-reasoning
- Direction
- Higher is better
- Sample
- 100
- Source type
- community-reported
- Evaluator
- DocPuzzle: A Process-Aware Benchmark for Evaluating Realistic Long-Context Reasoning Capabilities
- Status
- disputed
- Retrieved at
- 2026-10-18
- Methodology
- Community-reported result. Exact Table 3 row DeepSeek-R1-Distill-Qwen-32B. The paper reports temperature 0.3, top_p 0.2, top_k 0 for this model; GPT-4o judge; majority voting across three similar evaluation prompts; and an average over three evaluation-script runs. The exact endpoint is not stated.
- Context
- The paper's abstract says DeepSeek-R1-Distill-Qwen-32B scores 41.3%, while Table 3 assigns 39.7% to that model and 41.3% to QwQ-32B-Preview. The table is the only unambiguous model-specific row, but the source-internal discrepancy is retained as disputed. Total=100; three script runs average the same task set and do not imply N=300. No run-level values or confidence interval are published.
GPQA Diamond accuracy, FFS (Show measurement, test conditions, and source)
- Source value
- 62.6
- Score
- 62.6
- Metric
- accuracy (exact-answer match)
- Unit
- %
- Benchmark version
- GPQA Diamond (revision not stated)
- Direction
- Higher is better
- Sample
- 198
- Source type
- independent-evaluator
- Evaluator
- Agarwal et al., First Finish Search: Efficient Test-Time Scaling in Large Language Models, arXiv:2505.18149v1
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. Agarwal et al. report a zero-shot evaluation without additional fine-tuning, using First Finish Search (FFS) with n=4 independent stochastic samples per problem, beam size 1, top_p=0.95, temperature=0.6, and the DeepInfra API. For GPQA Diamond, maximum generation length is 16K tokens; a response is correct when the selected option exactly matches the answer.
- Context
- No confidence interval or immutable provider-side served revision is reported; the paper says GPQA Diamond comprises 198 questions.
AIME24 accuracy, FFS (Show measurement, test conditions, and source)
- Source value
- 80
- Score
- 80
- Metric
- accuracy (exact-answer match)
- Unit
- %
- Benchmark version
- AIME24 (revision not stated)
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Agarwal et al., First Finish Search: Efficient Test-Time Scaling in Large Language Models, arXiv:2505.18149v1
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. Agarwal et al. report a zero-shot evaluation without additional fine-tuning, using First Finish Search (FFS) with n=4 independent stochastic samples per problem, beam size 1, top_p=0.95, temperature=0.6, and the DeepInfra API. For AIME24, maximum generation length is 32K tokens; figures were removed and problems converted to text, and scoring checks the final integer answer.
- Context
- The paper says AIME24, AIME25-I and AIME25-II total 60 problems, but does not report per-dataset N. No confidence interval or immutable provider-side served revision is reported.
AIME25-I accuracy, FFS (Show measurement, test conditions, and source)
- Source value
- 59.6
- Score
- 59.6
- Metric
- accuracy (exact-answer match)
- Unit
- %
- Benchmark version
- AIME25-I (revision not stated)
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Agarwal et al., First Finish Search: Efficient Test-Time Scaling in Large Language Models, arXiv:2505.18149v1
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. Agarwal et al. report a zero-shot evaluation without additional fine-tuning, using First Finish Search (FFS) with n=4 independent stochastic samples per problem, beam size 1, top_p=0.95, temperature=0.6, and the DeepInfra API. For AIME25-I, maximum generation length is 32K tokens; figures were removed and problems converted to text, and scoring checks the final integer answer.
- Context
- The paper says AIME24, AIME25-I and AIME25-II total 60 problems, but does not report per-dataset N. No confidence interval or immutable provider-side served revision is reported.
AIME25-II accuracy, FFS (Show measurement, test conditions, and source)
- Source value
- 46.7
- Score
- 46.7
- Metric
- accuracy (exact-answer match)
- Unit
- %
- Benchmark version
- AIME25-II (revision not stated)
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Agarwal et al., First Finish Search: Efficient Test-Time Scaling in Large Language Models, arXiv:2505.18149v1
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. Agarwal et al. report a zero-shot evaluation without additional fine-tuning, using First Finish Search (FFS) with n=4 independent stochastic samples per problem, beam size 1, top_p=0.95, temperature=0.6, and the DeepInfra API. For AIME25-II, maximum generation length is 32K tokens; figures were removed and problems converted to text, and scoring checks the final integer answer.
- Context
- The paper says AIME24, AIME25-I and AIME25-II total 60 problems, but does not report per-dataset N. No confidence interval or immutable provider-side served revision is reported.
LiveCodeBench v2 Pass@1, DeepSeek-R1-Distill-Qwen-32B, Zero-Shot (Show measurement, test conditions, and source)
- Source value
- 69.1
- Score
- 69.1
- Metric
- Pass@1
- Unit
- %
- Benchmark version
- LiveCodeBench v2
- Category
- code generation
- Direction
- Higher is better
- Sample
- 511
- Source type
- independent-evaluator
- Evaluator
- Dacheng Li et al., S*: Test Time Scaling for Code Generation (arXiv:2502.14382v1)
- Status
- active
- Retrieved at
- 2026-10-21
- Methodology
- Independent evaluation. Baseline without test-time scaling (table row “Zero-Shot”). LiveCodeBench v2, 511 tasks; Pass@1.
- Context
- The paper reports one run per experiment. It gives no exact run date, immutable weights hash, or confidence interval.
LiveCodeBench v2 Pass@1, DeepSeek-R1-Distill-Qwen-32B, Majority Vote (Show measurement, test conditions, and source)
- Source value
- 75.8
- Score
- 75.8
- Metric
- Pass@1
- Unit
- %
- Benchmark version
- LiveCodeBench v2
- Category
- code generation
- Direction
- Higher is better
- Sample
- 511
- Source type
- independent-evaluator
- Evaluator
- Dacheng Li et al., S*: Test Time Scaling for Code Generation (arXiv:2502.14382v1)
- Status
- active
- Retrieved at
- 2026-10-21
- Methodology
- Independent evaluation. 16 parallel samples, clustered by execution results, then a random sample is selected from the largest cluster. GPT-4o mini generates clustering inputs for reasoning models. LiveCodeBench v2, 511 tasks; Pass@1.
- Context
- The paper reports one run per experiment. It gives no exact run date, immutable weights hash, or confidence interval.
LiveCodeBench v2 Pass@1, DeepSeek-R1-Distill-Qwen-32B, Self-Debugging (Show measurement, test conditions, and source)
- Source value
- 70.1
- Score
- 70.1
- Metric
- Pass@1
- Unit
- %
- Benchmark version
- LiveCodeBench v2
- Category
- code generation
- Direction
- Higher is better
- Sample
- 511
- Source type
- independent-evaluator
- Evaluator
- Dacheng Li et al., S*: Test Time Scaling for Code Generation (arXiv:2502.14382v1)
- Status
- active
- Retrieved at
- 2026-10-21
- Methodology
- Independent evaluation. One sample is iteratively refined with public tests for two rounds; for reasoning models, the model itself performs the refinement. LiveCodeBench v2, 511 tasks; Pass@1.
- Context
- The paper reports one run per experiment. It gives no exact run date, immutable weights hash, or confidence interval.
LiveCodeBench v2 Pass@1, DeepSeek-R1-Distill-Qwen-32B, S* (Show measurement, test conditions, and source)
- Source value
- 85.7
- Score
- 85.7
- Metric
- Pass@1
- Unit
- %
- Benchmark version
- LiveCodeBench v2
- Category
- code generation
- Direction
- Higher is better
- Sample
- 511
- Source type
- independent-evaluator
- Evaluator
- Dacheng Li et al., S*: Test Time Scaling for Code Generation (arXiv:2502.14382v1)
- Status
- active
- Retrieved at
- 2026-10-21
- Methodology
- Independent evaluation. 16 parallel samples at temperature 0.7 with no top-p sampling; two rounds of iterative debugging on public tests and adaptive input generation for selection. LiveCodeBench v2, 511 tasks; Pass@1.
- Context
- The paper reports one run per experiment. It gives no exact run date, immutable weights hash, or confidence interval.
Page 1 of 6
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 7, 2026
DeepSeek-R1-Distill-Qwen-32BMagistral Small 53Nemotron 3.5 Super VL 120B A12B 53Spark-X2.5-4B 52Gemini 3.7 Flash 52
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 DeepSeek-R1-Distill-Qwen-32B, chosen by model line, developer, size, price and release date.
- DeepSeek-R1-Distill-Llama-70BDeepSeek · same developer
- DeepSeek R1 0528 Qwen3 8BDeepSeek · same developer
- DeepSeek LLM 67BDeepSeek · same developer
- DeepSeek-R1DeepSeek · same developer
- DeepSeek-V3DeepSeek · same developer
- DeepSeek-V3 0324DeepSeek · same developer
- DeepSeek Coder V2DeepSeek · same developer
- DeepSeek-R1 0528DeepSeek · 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 DeepSeek-R1-Distill-Qwen-32B
How large is the context window of DeepSeek-R1-Distill-Qwen-32B?
131,072 tokens. That is how much text DeepSeek-R1-Distill-Qwen-32B reads at once, about 200 A4 pages.
Page count as a rule of thumb from Artificial Analysis
GLM-5.1200K
DeepSeek-V3 0324164K
DeepSeek-R1-Distill-Qwen-32BRank 94 of 103131K
GLM-4.5131K
GPT OSS 120B131K
DeepSeek-V3.2128K
DeepSeek-R1128K
Can I run DeepSeek-R1-Distill-Qwen-32B locally?
Yes, the weights are open.
Stated by DeepSeek
- Open weights
- yes
- License
- MIT
- Parameters
- 32B
- active per token
- n/a
Sources and data date
Show all 27 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
| Evaluator | Test series | Composite indices | Recorded active measurements |
|---|---|---|---|
| Artificial Analysis | 1 | 0 | 1 |
Without vendor claims and historical or disputed results; those stay labelled in the single values.
| Type | Evidence and data date |
|---|---|
| Input | Documented modalities. Retrieved October 27, 2026. |
| Output | Documented modalities. Retrieved October 27, 2026. |
| Research status | Research date October 15, 2026. 18 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not published | task-snapshots Note The card does not state exact run dates or task snapshots for MATH-500, GPQA Diamond, and LiveCodeBench. |
| Not published | codeforces-rating-protocol Note The card publishes a 1691 CodeForces rating but not the contest snapshot or rating procedure. |
| Not published | immutable served model revision and exact endpoint Note The added source row names the model but does not publish an immutable served revision or, where applicable, exact provider endpoint. No result is transferred to a dated sibling variant. |
| Not published | unreported uncertainty metadata Note Unreported confidence intervals and per-subcategory sample counts remain absent; displayed ± terms with undefined meaning are not converted to intervals. |
| Additional source | Hugging Face (retrieved October 15, 2026) · DeepSeek-R1-Distill-Qwen-32B model card · Editorial description reviewed October 2, 2026 |
| Additional source | Artificial Analysis (retrieved October 19, 2026; November 2, 2026; November 7, 2026) |
| Additional source | PsyCrisisBench authors, arXiv v1 Supplementary Table 4 (retrieved October 17, 2026) |
| Additional source | DocPuzzle: A Process-Aware Benchmark for Evaluating Realistic Long-Context Reasoning Capabilities (retrieved October 18, 2026) |
| Additional source | Agarwal et al., First Finish Search: Efficient Test-Time Scaling in Large Language Models, arXiv:2505.18149v1 (retrieved October 19, 2026) |
| Additional source | Dacheng Li et al., S*: Test Time Scaling for Code Generation (arXiv:2502.14382v1) (retrieved October 21, 2026) |
| Additional source | Wu et al., OPV: Outcome-based Process Verifier for Efficient Long Chain-of-Thought Verification (arXiv:2512.10756v1) (retrieved October 21, 2026) |
| Additional source | BALROG team (retrieved October 22, 2026) |
| Additional source | Epoch AI (retrieved October 22, 2026) |
| Additional source | TimE: A Multi-level Benchmark for Temporal Reasoning of LLMs in Real-World Scenarios (arXiv:2505.12891v1) (retrieved October 22, 2026) |
| Additional source | BigCodeBench official leaderboard (retrieved October 28, 2026)12 |
| Additional source | Model metadata (retrieved October 17, 2026)123 |
| Additional source | Model metadata (retrieved October 28, 2026)1234 |
| Additional source | DeepSeek-R1 distillation documentation · 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 | Ollama: cumulative downloads for the popularity index |
| Additional source | Replicate: cumulative runs for the popularity index |
| Additional source | Google Trending Now: limited feed selection for Germany and the US |
| Additional source | Creative Commons Attribution 4.0 |