# Image Classification on ImageNet-21K Source: [Image Classification on ImageNet-21K]() Collection index: [llms.txt]() Data updated: 2026-05-26T06:45:30.684Z Evaluation of standard Image Classification performance on the ImageNet-21K dataset, measured using Top-1 accuracy. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Image Classification | | Dataset (including any stated split) | ImageNet-21K | | Metric | Top-1 Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 25 of 25 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Top-1 Accuracy](): Higher is better; unit: PERCENT. - [Absolute Improvement](): Higher is better; unit: SCALAR. ## Recorded results | Method | Top-1 Accuracy | Result date | Report paper | | --- | --- | --- | --- | | NomMer-B | 86.6 | Not recorded | 68fa6bfe1e209131cea3186a | | Swin-B | 86.4 | Not recorded | 68fa6bfe1e209131cea3186a | | ViL-B | 86.2 | Not recorded | 68fa6bfe1e209131cea3186a | | NomMer-B | 85.5 | Not recorded | 68fa6bfe1e209131cea3186a | | Swin-B | 85.2 | Not recorded | 68fa6bfe1e209131cea3186a | | R-101x3 | 84.4 | Not recorded | 68fa6bfe1e209131cea3186a | | ViT-B/16 | 84 | Not recorded | 68fa6bfe1e209131cea3186a | | Claude Opus 4.5 | 61.64 | Not recorded | 6a15290fdbd7150506f7de45 | | Claude Opus 4.5 | 45.48 | Not recorded | 6a15290fdbd7150506f7de45 | | GPT-4(v4.1) | 45.1 | Not recorded | 6a15290fdbd7150506f7de45 | | Qwen3-VL-Plus | 42.2 | Not recorded | 6a15290fdbd7150506f7de45 | | Qwen2.5-VL-7B | 37.87 | Not recorded | 6a15290fdbd7150506f7de45 | | GPT-4(v4.1) | 36.36 | Not recorded | 6a15290fdbd7150506f7de45 | | Qwen3-VL-4B | 34.41 | Not recorded | 6a15290fdbd7150506f7de45 | | Qwen3-VL-8B | 34.2 | Not recorded | 6a15290fdbd7150506f7de45 | | Kimi-K2.5 | 32.26 | Not recorded | 6a15290fdbd7150506f7de45 | | Qwen3-VL-2B | 19.29 | Not recorded | 6a15290fdbd7150506f7de45 | | Kimi-K2.5 | 15.82 | Not recorded | 6a15290fdbd7150506f7de45 | | LLaMA-4 | 14.28 | Not recorded | 6a15290fdbd7150506f7de45 | | LLaMA-4 | 11.14 | Not recorded | 6a15290fdbd7150506f7de45 | | Qwen3-VL-Plus | 10.56 | Not recorded | 6a15290fdbd7150506f7de45 | | Qwen3-VL-8B | 10.56 | Not recorded | 6a15290fdbd7150506f7de45 | | Qwen3-VL-4B | 0.96 | Not recorded | 6a15290fdbd7150506f7de45 | | Qwen3-VL-2B | 0.48 | Not recorded | 6a15290fdbd7150506f7de45 | | Qwen2.5-VL-7B | 0.21 | Not recorded | 6a15290fdbd7150506f7de45 | ### NomMer-B Top-1 Accuracy: 86.6. Result ID: 698eb644ffc3901de0a00262. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Resolution":"384x384","#param. (M)":"73","FLOPs (G)":"56.2"}. This record does not assert independent reproduction. ### Swin-B Top-1 Accuracy: 86.4. Result ID: 698eb644ffc3901de0a00261. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Resolution":"384x384","#param. (M)":"88","FLOPs (G)":"47.1"}. This record does not assert independent reproduction. ### ViL-B Top-1 Accuracy: 86.2. Result ID: 698eb644ffc3901de0a00260. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Resolution":"384x384","#param. (M)":"56","FLOPs (G)":"43.7"}. This record does not assert independent reproduction. ### NomMer-B Top-1 Accuracy: 85.5. Result ID: 698eb644ffc3901de0a0025d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Resolution":"224x224","#param. (M)":"73","FLOPs (G)":"17.6"}. This record does not assert independent reproduction. ### Swin-B Top-1 Accuracy: 85.2. Result ID: 698eb644ffc3901de0a0025c. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Resolution":"224x224","#param. (M)":"88","FLOPs (G)":"15.4"}. This record does not assert independent reproduction. ### R-101x3 Top-1 Accuracy: 84.4. Result ID: 698eb644ffc3901de0a0025e. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Resolution":"384x384","#param. (M)":"388","FLOPs (G)":"204.6"}. This record does not assert independent reproduction. ### ViT-B/16 Top-1 Accuracy: 84. Result ID: 698eb644ffc3901de0a0025f. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Resolution":"384x384","#param. (M)":"86","FLOPs (G)":"55.4"}. This record does not assert independent reproduction. ### Claude Opus 4.5 Top-1 Accuracy: 61.64. Result ID: 6a15416efbc373592c0c01e5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Divide-and-Conquer Inference (DCI)","K":"100"}. This record does not assert independent reproduction. ### Claude Opus 4.5 Top-1 Accuracy: 45.48. Result ID: 6a15416efbc373592c0c01e4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Baseline"}. This record does not assert independent reproduction. ### GPT-4(v4.1) Top-1 Accuracy: 45.1. Result ID: 6a15416efbc373592c0c01e3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Divide-and-Conquer Inference (DCI)","K":"100"}. This record does not assert independent reproduction. ### Qwen3-VL-Plus Top-1 Accuracy: 42.2. Result ID: 6a15416efbc373592c0c01e7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Divide-and-Conquer Inference (DCI)","K":"100"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Top-1 Accuracy: 37.87. Result ID: 6a15416efbc373592c0c01ed. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Divide-and-Conquer Inference (DCI)","K":"100"}. This record does not assert independent reproduction. ### GPT-4(v4.1) Top-1 Accuracy: 36.36. Result ID: 6a15416efbc373592c0c01e2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Baseline"}. This record does not assert independent reproduction. ### Qwen3-VL-4B Top-1 Accuracy: 34.41. Result ID: 6a15416efbc373592c0c01f1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Divide-and-Conquer Inference (DCI)","K":"100"}. This record does not assert independent reproduction. ### Qwen3-VL-8B Top-1 Accuracy: 34.2. Result ID: 6a15416efbc373592c0c01f3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Divide-and-Conquer Inference (DCI)","K":"100"}. This record does not assert independent reproduction. ### Kimi-K2.5 Top-1 Accuracy: 32.26. Result ID: 6a15416efbc373592c0c01e9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Divide-and-Conquer Inference (DCI)","K":"100"}. This record does not assert independent reproduction. ### Qwen3-VL-2B Top-1 Accuracy: 19.29. Result ID: 6a15416efbc373592c0c01ef. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Divide-and-Conquer Inference (DCI)","K":"100"}. This record does not assert independent reproduction. ### Kimi-K2.5 Top-1 Accuracy: 15.82. Result ID: 6a15416efbc373592c0c01e8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Baseline"}. This record does not assert independent reproduction. ### LLaMA-4 Top-1 Accuracy: 14.28. Result ID: 6a15416efbc373592c0c01eb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Divide-and-Conquer Inference (DCI)","K":"100"}. This record does not assert independent reproduction. ### LLaMA-4 Top-1 Accuracy: 11.14. Result ID: 6a15416efbc373592c0c01ea. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Baseline"}. This record does not assert independent reproduction. ### Qwen3-VL-Plus Top-1 Accuracy: 10.56. Result ID: 6a15416efbc373592c0c01e6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Baseline"}. This record does not assert independent reproduction. ### Qwen3-VL-8B Top-1 Accuracy: 10.56. Result ID: 6a15416efbc373592c0c01f2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Baseline"}. This record does not assert independent reproduction. ### Qwen3-VL-4B Top-1 Accuracy: 0.96. Result ID: 6a15416efbc373592c0c01f0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Baseline"}. This record does not assert independent reproduction. ### Qwen3-VL-2B Top-1 Accuracy: 0.48. Result ID: 6a15416efbc373592c0c01ee. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Baseline"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Top-1 Accuracy: 0.21. Result ID: 6a15416efbc373592c0c01ec. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Strategy":"Baseline"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.