INaturalist Unveils Major AI Upgrade: Computer Vision Model v.2.31 Expands to 118,700 Taxa

By Science & Technology Desk
May 20, 2026

In a significant leap forward for global biodiversity monitoring, iNaturalist has officially released its latest computer vision and geomodel update, v.2.31. This iteration marks a pivotal moment in the platform’s mission to digitize the natural world, bringing the total number of recognized taxa to 118,700. This expansion, powered by the collective efforts of millions of citizen scientists, represents a refined, more accurate tool for researchers, naturalists, and casual observers alike.

Main Facts: The Power of Collective Intelligence

The v.2.31 model, which was trained on data exported as of April 12, 2026, represents a substantial increase from the previous v.2.30 model, which supported 117,318 taxa. This growth is not merely a statistical figure; it is a testament to the surging volume of high-quality, research-grade data being contributed to the iNaturalist ecosystem daily.

At its core, the computer vision model serves as a digital taxonomic assistant. When a user uploads a photo of a plant, insect, or animal, the model analyzes the visual features and geographical metadata to suggest potential identifications. To be included in the model, a taxon must meet rigorous thresholds: typically, at least 100 verified photographs and 60 distinct observations are required to ensure the AI has sufficient data to "learn" the unique morphological characteristics of that species, genus, or family.

This update underscores the symbiotic relationship between human expertise and machine learning. While the AI suggests identifications, it is the community of human identifiers—expert taxonomists, hobbyists, and students—who curate the data. As these human experts correct misidentifications and update taxonomic classifications, the AI is "retrained" to reflect the most current scientific consensus.

Chronology: A Trajectory of Exponential Growth

To understand the significance of this update, one must view it within the broader historical context of iNaturalist’s technological evolution. Over the past four years, the platform has seen an explosive increase in its capacity to identify life on Earth.

In 2022, the model supported approximately 55,000 taxa. By mid-2026, that number has more than doubled. This rapid scaling is the result of a deliberate, iterative development cycle. The iNaturalist team currently aims to update the model every one to two months. This high frequency is essential for several reasons:

1. Taxonomic Fluidity: Scientific understanding of phylogeny is constantly evolving. As DNA sequencing and morphological research reveal new species or reclassify old ones, the model must be updated to maintain scientific accuracy.
2. The Feedback Loop: By providing monthly updates, the platform ensures that the community’s recent contributions—especially those involving rare or newly documented species—are integrated into the tool as quickly as possible.
3. Algorithmic Refinement: Each release is not just an addition of new species; it is an optimization of the underlying neural network, allowing the system to distinguish between look-alike species with greater precision.

The steady climb in the number of taxa reflects a broader trend of increased global engagement in citizen science, facilitated by accessible mobile technology and a growing public awareness of the biodiversity crisis.

Supporting Data: Validating Accuracy

A common concern in AI development is the trade-off between the breadth of coverage and the precision of results. Adding more taxa could theoretically lead to "model fatigue," where the system becomes less accurate due to the complexity of distinguishing between a higher number of similar-looking species.

However, the data for the v.2.31 update indicates a positive trend. To evaluate the new model, the iNaturalist team performs a comparative analysis against the previous version using a rigorous methodology. They utilize a set of 1,000 random, Research Grade observations for each taxonomic group—observations that were deliberately excluded from the training data to avoid bias.

The results, visualized in comparative performance charts, demonstrate that the v.2.31 model maintains or exceeds the accuracy of its predecessor. By comparing the average accuracy of v.2.30 against v.2.31, the developers have confirmed that the model’s ability to correctly categorize life has kept pace with its expansion. This is critical for users who rely on these suggestions to build their own knowledge or to contribute to scientific datasets where misidentification would be detrimental.

Official Responses and Community Impact

The release of v.2.31 was met with enthusiasm from the community, with key contributors and staff acknowledging the collaborative nature of this milestone. As noted by the iNaturalist team, the model is a collective work of art—a synthesis of millions of individual decisions made by community members who have painstakingly tagged, identified, and verified images from every corner of the globe.

"Thank you to everyone in the community who contributed observations and identifications for all the species in this model," the team stated in the release announcement. "This collective effort wouldn’t be possible without you."

The platform also maintains a dedicated help page that provides transparency regarding how taxa are selected for the model. This transparency is vital for maintaining the trust of the scientific community. By detailing the specific requirements for inclusion, iNaturalist encourages users to focus their efforts on documenting species that are currently underrepresented, thereby "training" the model to be more inclusive of diverse and rare ecosystems.

Implications: The Future of Biodiversity Informatics

The implications of a 118,700-taxon model are profound. As the world faces unprecedented rates of habitat loss and climate change, the ability to rapidly identify species is a fundamental requirement for conservation planning.

Accelerating Scientific Discovery

For biologists, the AI acts as a force multiplier. It allows researchers to process thousands of images in a fraction of the time it would take to manually sort through them. This is particularly valuable for "BioBlitz" events or long-term ecological monitoring projects where volume is high.

Bridging the Knowledge Gap

The accessibility of this tool on a smartphone democratizes science. A student in a remote area with little to no formal training in taxonomy can now access an identification tool that rivals the knowledge base of a seasoned naturalist. This encourages participation in local environmental monitoring, potentially leading to the discovery of invasive species or the reappearance of rare, presumed-extinct taxa.

The Challenge of "Machine-Assisted" Identification

As the model grows more sophisticated, the role of the human identifier changes. We are moving toward a paradigm where the AI performs the first pass of identification, and the human expert acts as the final arbiter. This division of labor is essential for managing the scale of data that iNaturalist receives. However, it also places a premium on the quality of human input. The integrity of the model depends on the accuracy of the identifications made by users. If users blindly accept AI suggestions without verifying them, the model risks reinforcing its own errors.

Looking Ahead

The path forward for iNaturalist involves continued integration of finer-grained data, including environmental layers and sound-based identification. The success of v.2.31 serves as a roadmap for what is possible when data-driven technology is paired with a passionate, globally distributed community of volunteers.

As we look toward future updates, the goal remains clear: to provide the most accurate, accessible, and comprehensive window into the natural world. With each update, the "digital twin" of our planet’s biodiversity becomes more detailed, providing an invaluable record for current researchers and a foundational legacy for future generations.

The release of v.2.31 is a milestone that honors the past efforts of the iNaturalist community while setting a high bar for the future. Whether you are an expert botanist or a curious observer taking your first photograph, you are now a participant in the most comprehensive taxonomic database in human history. The next time you open the iNaturalist app, remember that the suggestion you receive is the result of over 118,000 species’ worth of collective wisdom—a digital testament to the beauty and complexity of life on Earth.