Benchmark Almanac

Natural World Tasks (NeWT 2021)

BioCLIP-2.5 ViT-H/14 90.2% tops its leaderboard
Not in model cards

Models distinguish pairs of labels in wildlife photographs. The 164 tasks include recognizing life stages, animal behavior and scene context.

Passes test 4. The BioCLIP 2.5 model card reports accuracy across test images. The original paper averages accuracy across tasks.

Released
March 2021
Kind
General visual questions
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
English
In and out
Images, Text → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
59
Entries on BioBench’s leaderboard, 0 models we track

BioCLIP-2.5 ViT-H/14 tops the leaderboard with 90.2%

Best score per entry from BioBench’s leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

  1. 1 BioCLIP-2.5 ViT-H/14 90.2%
  2. 2 PE ViT-G/14 89.5%
  3. 3 BioCLIP-2 ViT-L/14 89.1%
  4. 4 PE ViT-L/14 87.8%
  5. 5 SigLIP2 ViT-1B/16 86.7%
  6. 6 SigLIP SO400M/14 86%
  7. 7 SigLIP2 SO400M/16 86%
  8. 8 DINOv3 ViT-L/16 85.6%
  9. 9 SigLIP ViT-L/16 85.4%
  10. 10 DINOv3 ViT-B/16 85.2%
  11. 11 SigLIP2 ViT-L/16 85%
  12. 12 MetaCLIP 2 ViT-H/14 84.8%
  13. 13 BioCAP ViT-B/16 84.7%
  14. 14 PE ViT-B/16 84.6%
  15. 15 SigLIP2 ViT-B/16 84.4%
  16. 16 AIMv2 ViT-1B/14 84.2%
  17. 17 SigLIP ViT-B/16 84.2%
  18. 18 AIMv2 ViT-3B/14 84%
  19. 19 OpenVision SO400M/14 83.8%
  20. 20 CLIP ViT-L/14 83.6%
  21. 21 OpenVision ViT-H/14 83.6%
  22. 22 DINOv2 ViT-B/14 83%
  23. 23 OpenVision ViT-L/14 83%
  24. 24 DINOv2 ViT-L/14 83%
  25. 25 DINOv3 ViT-S/16 82.9%
  26. 26 BioTrove ViT-B/16 82.9%
  27. 27 DINOv2 ViT-g/14 82.8%
  28. 28 BioCLIP ViT-B/16 82.7%
  29. 29 PE ViT-S/16 82.5%
  30. 30 AIMv2 ViT-H/14 82.3%
  31. 31 DINOv2 ViT-S/14 82%
  32. 32 ViT-B/16 (IN1K) 81.7%
  33. 33 OpenVision ViT-B/16 81.1%
  34. 34 ConvNext-T 80.1%
  35. 35 CLIP ViT-B/16 80%
  36. 36 AIMv2 ViT-L/14 79.8%
  37. 37 PE Spatial ViT-G/14 79.3%
  38. 38 OpenVision ViT-S/16 78.6%
  39. 39 SigLIP2 ViT-B/32 78.4%
  40. 40 CLIP ViT-B/32 78%
  41. 41 ConvNext-T (IN1K) 77.1%
  42. 42 OpenVision ViT-T/16 74.5%
  43. 43 MegaDescriptor Swin-L/4 71.3%
  44. 44 MobileNet V3 L 69.8%
  45. 45 ResNet-18 69%
  46. 46 EfficientNet B0 68.7%
  47. 47 V-JEPA ViT-H/16 68%
  48. 48 ResNet-50 67.8%
  49. 49 V-JEPA ViT-L/16 67.4%
  50. 50 MobileNet V2 67.3%
  51. 51 PE Spatial ViT-B/16 67%
  52. 52 ResNet-50d 66.2%
  53. 53 EfficientNet B3 66.1%
  54. 54 SAM2 Hiera-T 66.1%
  55. 55 PE Spatial ViT-L/14 65.7%
  56. 56 SAM2 Hiera-S 65.1%
  57. 57 SAM2 Hiera-L 64.2%
  58. 58 SAM2 Hiera-B+ 64.1%
  59. 59 PE Spatial ViT-T/16 64%

Source: BioBench’s leaderboard

#EntryScore ±SettingDateSource
1 BioCLIP-2.5 ViT-H/14 0.6B 90.2% ±0.4 Official leaderboard
2 PE ViT-G/14 1.9B 89.5% 89 to 90 Official leaderboard
3 BioCLIP-2 ViT-L/14 0.3B 89.1% 88.7 to 89.6 Official leaderboard
4 PE ViT-L/14 0.3B 87.8% ±0.5 Official leaderboard
5 SigLIP2 ViT-1B/16 1.2B 86.7% ±0.5 Official leaderboard
6 SigLIP SO400M/14 0.4B 86% ±0.5 Official leaderboard
7 SigLIP2 SO400M/16 0.4B 86% ±0.5 Official leaderboard
8 DINOv3 ViT-L/16 0.3B 85.6% ±0.5 Official leaderboard
9 SigLIP ViT-L/16 0.3B 85.4% 84.9 to 85.8 Official leaderboard
10 DINOv3 ViT-B/16 0.1B 85.2% 84.7 to 85.7 Official leaderboard
11 SigLIP2 ViT-L/16 0.3B 85% ±0.5 Official leaderboard
12 MetaCLIP 2 ViT-H/14 0.6B 84.8% 84.3 to 85.3 Official leaderboard
13 BioCAP ViT-B/16 0.1B 84.7% 84.1 to 85.2 Official leaderboard
14 PE ViT-B/16 0.1B 84.6% 84.1 to 85.1 Official leaderboard
15 SigLIP2 ViT-B/16 0.1B 84.4% 83.9 to 84.9 Official leaderboard
16 AIMv2 ViT-1B/14 1.2B 84.2% ±0.5 Official leaderboard
17 SigLIP ViT-B/16 0.1B 84.2% ±0.5 Official leaderboard
18 AIMv2 ViT-3B/14 2.7B 84% 83.5 to 84.5 Official leaderboard
19 OpenVision SO400M/14 0.4B 83.8% ±0.6 Official leaderboard
20 CLIP ViT-L/14 0.3B 83.6% ±0.6 Official leaderboard
21 OpenVision ViT-H/14 0.6B 83.6% 83 to 84.1 Official leaderboard
22 DINOv2 ViT-B/14 0.1B 83% 82.5 to 83.6 Official leaderboard
23 OpenVision ViT-L/14 0.3B 83% ±0.6 Official leaderboard
24 DINOv2 ViT-L/14 0.3B 83% ±0.5 Official leaderboard
25 DINOv3 ViT-S/16 0B 82.9% ±0.6 Official leaderboard
26 BioTrove ViT-B/16 0.1B 82.9% ±0.6 Official leaderboard
27 DINOv2 ViT-g/14 1.1B 82.8% 82.3 to 83.4 Official leaderboard
28 BioCLIP ViT-B/16 0.1B 82.7% ±0.5 Official leaderboard
29 PE ViT-S/16 0B 82.5% 81.9 to 83 Official leaderboard
30 AIMv2 ViT-H/14 0.7B 82.3% ±0.5 Official leaderboard
31 DINOv2 ViT-S/14 0B 82% 81.5 to 82.6 Official leaderboard
32 ViT-B/16 (IN1K) 0.1B 81.7% 81.1 to 82.3 Official leaderboard
33 OpenVision ViT-B/16 0.1B 81.1% 80.5 to 81.7 Official leaderboard
34 ConvNext-T 0B 80.1% 79.5 to 80.7 Official leaderboard
35 CLIP ViT-B/16 0.1B 80% ±0.6 Official leaderboard
36 AIMv2 ViT-L/14 0.3B 79.8% ±0.6 Official leaderboard
37 PE Spatial ViT-G/14 1.9B 79.3% ±0.6 Official leaderboard
38 OpenVision ViT-S/16 0B 78.6% 78 to 79.2 Official leaderboard
39 SigLIP2 ViT-B/32 0.1B 78.4% 77.8 to 79 Official leaderboard
40 CLIP ViT-B/32 0.1B 78% 77.5 to 78.6 Official leaderboard
41 ConvNext-T (IN1K) 0B 77.1% 76.5 to 77.7 Official leaderboard
42 OpenVision ViT-T/16 0B 74.5% ±0.7 Official leaderboard
43 MegaDescriptor Swin-L/4 0.2B 71.3% 70.7 to 72 Official leaderboard
44 MobileNet V3 L 0B 69.8% 69.1 to 70.3 Official leaderboard
45 ResNet-18 0B 69% 68.4 to 69.7 Official leaderboard
46 EfficientNet B0 0B 68.7% ±0.7 Official leaderboard
47 V-JEPA ViT-H/16 0.6B 68% ±0.7 Official leaderboard
48 ResNet-50 0B 67.8% 67.1 to 68.5 Official leaderboard
49 V-JEPA ViT-L/16 0.3B 67.4% ±0.7 Official leaderboard
50 MobileNet V2 0B 67.3% ±0.7 Official leaderboard
51 PE Spatial ViT-B/16 0.1B 67% ±0.7 Official leaderboard
52 ResNet-50d 0B 66.2% 65.5 to 66.7 Official leaderboard
53 EfficientNet B3 0B 66.1% ±0.7 Official leaderboard
54 SAM2 Hiera-T 0B 66.1% 65.4 to 66.8 Official leaderboard
55 PE Spatial ViT-L/14 0.3B 65.7% 64.9 to 66.3 Official leaderboard
56 SAM2 Hiera-S 0B 65.1% 64.4 to 65.8 Official leaderboard
57 SAM2 Hiera-L 0.2B 64.2% 63.5 to 64.8 Official leaderboard
58 SAM2 Hiera-B+ 0.1B 64.1% 63.4 to 64.8 Official leaderboard
59 PE Spatial ViT-T/16 0B 64% ±0.7 Official leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Vision

All 636 in Vision
  • MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy. Qwen 3.6 Plus 86%
  • MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
  • CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
  • MathVista In use Math reasoning in visual contexts (charts, geometry, puzzles) from 31 datasets. Answer accuracy. Seed2.1 Pro 90.7%
  • MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy. Qwen 3.8 Max 97.7%
  • MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy. Qwen 3.8 Max 85.9%

Data audit

All 8 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the official leaderboard (59 entries).
  • Pass Models scored 59 entries on its leaderboard, none a model we track.
  • Note Newest score No scores yet.
  • Note Frontier models No models scored yet.
  • Note Headroom No scores yet.
Leaderboard import 3 pass
  • Pass Importer We read the board on BioBench: 59 entries at the last read, none a model we track.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check