DeepSeek: DeepSeek-VL2
DeepSeek-VL2 is a mixed-expert (MoE) vision-language model developed based on DeepSeekMoE-27B, employing a sparse-activated MoE architecture to achieve superior performance with only 4.5B active para…
siliconFlow/deepseek-ai/deepseek-vl2
DeepSeek-VL2 is a mixed-expert (MoE) vision-language model developed based on DeepSeekMoE-27B, employing a sparse-activated MoE architecture to achieve superior performance with only 4.5B active parameters. The model excels in various tasks including visual question answering, optical character recognition, document/table/chart understanding, and visual grounding. Compared to existing open-source dense models and MoE-based models, it demonstrates competitive or state-of-the-art performance using the same or fewer active parameters.
Specifications & capabilities
- Input
- text, image
- Output
- text
- Max output · tokens
- 4096
- Published model ID
- deepseek-vl2
- Capabilities
- multimodal, structuredOutputs, visionInput
Evaluations
41–60 / 116
84.4%
- source id
- mmBench
- leaderboard name
- sketch_reasoning
- dimension
- vision
- metric
- sketch_reasoning
- rank
- 44
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- CCBench_Data
- verified
- Yes
- metric_kind
- sketch_reasoning
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 233
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗75%
- source id
- mmBench
- leaderboard name
- future_prediction
- dimension
- vision
- metric
- future_prediction
- rank
- 44
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_CN_Data
- verified
- Yes
- metric_kind
- future_prediction
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 163
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗69.9%
- source id
- mmBench
- leaderboard name
- LR
- dimension
- vision
- metric
- LR
- rank
- 46
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_CN_Data
- verified
- Yes
- metric_kind
- LR
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 236
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗89.6%
- source id
- mmBench
- leaderboard name
- image_scene
- dimension
- vision
- metric
- image_scene
- rank
- 47
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_CN_V11_Data
- verified
- Yes
- metric_kind
- image_scene
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 295
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗75%
- source id
- mmBench
- leaderboard name
- future_prediction
- dimension
- vision
- metric
- future_prediction
- rank
- 47
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_EN_Data
- verified
- Yes
- metric_kind
- future_prediction
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 162
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗31.4%
- source id
- mmBench
- leaderboard name
- historical_figure
- dimension
- vision
- metric
- historical_figure
- rank
- 48
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- CCBench_Data
- verified
- Yes
- metric_kind
- historical_figure
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 233
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗97.4%
- source id
- mmBench
- leaderboard name
- identity_reasoning
- dimension
- vision
- metric
- identity_reasoning
- rank
- 49
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_EN_V11_Data
- verified
- Yes
- metric_kind
- identity_reasoning
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 295
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗85.2%
- source id
- mmBench
- leaderboard name
- nature_relation
- dimension
- vision
- metric
- nature_relation
- rank
- 50
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_EN_Data
- verified
- Yes
- metric_kind
- nature_relation
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 162
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗81.3%
- source id
- mmBench
- leaderboard name
- attribute_comparison
- dimension
- vision
- metric
- attribute_comparison
- rank
- 50
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_CN_V11_Data
- verified
- Yes
- metric_kind
- attribute_comparison
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 295
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗87.7%
- source id
- mmBench
- leaderboard name
- FP-S
- dimension
- vision
- metric
- FP-S
- rank
- 51
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_EN_Data
- verified
- Yes
- metric_kind
- FP-S
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 235
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗84.7%
- source id
- mmBench
- leaderboard name
- celebrity_recognition
- dimension
- vision
- metric
- celebrity_recognition
- rank
- 51
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_CN_Data
- verified
- Yes
- metric_kind
- celebrity_recognition
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 163
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗60.5%
- source id
- mmBench
- leaderboard name
- image_quality
- dimension
- vision
- metric
- image_quality
- rank
- 51
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_CN_V11_Data
- verified
- Yes
- metric_kind
- image_quality
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 295
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗66.7%
- source id
- mmBench
- leaderboard name
- calligraphy_painting
- dimension
- vision
- metric
- calligraphy_painting
- rank
- 52
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- CCBench_Data
- verified
- Yes
- metric_kind
- calligraphy_painting
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 233
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗85.9%
- source id
- mmBench
- leaderboard name
- FP-S
- dimension
- vision
- metric
- FP-S
- rank
- 53
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_CN_Data
- verified
- Yes
- metric_kind
- FP-S
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 236
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗77.3%
- source id
- mmBench
- leaderboard name
- FP-C
- dimension
- vision
- metric
- FP-C
- rank
- 53
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_EN_Data
- verified
- Yes
- metric_kind
- FP-C
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 235
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗68.3%
- source id
- mmBench
- leaderboard name
- structuralized_imagetext_understanding
- dimension
- vision
- metric
- structuralized_imagetext_understanding
- rank
- 54
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_EN_Data
- verified
- Yes
- metric_kind
- structuralized_imagetext_understanding
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 162
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗94.8%
- source id
- mmBench
- leaderboard name
- ocr
- dimension
- vision
- metric
- ocr
- rank
- 55
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_EN_Data
- verified
- Yes
- metric_kind
- ocr
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 162
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗71.1%
- source id
- mmBench
- leaderboard name
- LR
- dimension
- vision
- metric
- LR
- rank
- 55
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_EN_Data
- verified
- Yes
- metric_kind
- LR
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 235
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗89.9%
- source id
- mmBench
- leaderboard name
- attribute_recognition
- dimension
- vision
- metric
- attribute_recognition
- rank
- 56
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_CN_Data
- verified
- Yes
- metric_kind
- attribute_recognition
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 163
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗85.9%
- source id
- mmBench
- leaderboard name
- social_relation
- dimension
- vision
- metric
- social_relation
- rank
- 57
- aggregation scope
- status
- published
- direction
- higherIsBetter
- conditions
- suite
- MMBench_TEST_EN_Data
- verified
- Yes
- metric_kind
- social_relation
- rank_method
- competition rank among reported evaluations in identical conditions, including unlinked observations
- source_file
- mmbench-data.json
- source_revision
- acb08c72ac038d24cf7785254678efe725354fe91091bfa10a667832ad964b24
- metric_population
- 162
- benchmark_evaluation
- true
- source model id
- DeepSeek-VL2
Original result ↗Provider specifications & access
1–1 / 1
- source model id
- deepseek-ai/deepseek-vl2
- hosting
- siliconflow-hosted
- max output tokens
- 4096
Safety assessments
1–1 / 1
32 / 100
- source id
- stanfordFMTI
- canonical model id
- siliconFlow/deepseek-ai/deepseek-vl2
- taxonomy
- transparency
- hazard category
- transparency
- hazard name
- Organization transparency
- policy area
- DeepSeek
- assessment type
- organization-level disclosure
- metric
- disclosure_score
- result label
- 32 / 100
- sample size
- 100
- updated at
- 2026-11-09T18:08:05.227Z
Source specification observations
1–2 / 2
- field
- context_length
- value
- 4K
- scope
- source-model
- source id
- siliconFlow
- confidence
- 1
- field
- max_output_tokens
- value
- 4K
- scope
- source-model
- source id
- siliconFlow
- confidence
- 1
Explore more models