New de novo design model from Baker Lab + NVIDIA, available on Tamarind today. We are excited to be working with @NVIDIAHealth to bring access to RFD4-Proteina on day 0.

RFD4-Proteina is available on @tamarindbio today, alongside hundreds of molecular design models. Tamarind also offers a streamlined workflow for fine-tuning the model on your own data. Try it out: app.tamarind.bio/tools/rfd4-…

BindCraft2 is out, ready to use on Tamarind on day one. The workflow expands on the original BindCraft substantially. Most notably, now supporting VHH, scFv, Fab, and seven other binder modalities, supporting multi-state targets, and multiple targets at once.

Claude speeds up molecular design models: 4x on average @AnthropicAI took open biomolecular modeling tools like ESMFold2, OpenDDE, and RFdiffusion, and let Claude (Science) optimize the model kernels, the part of the software interfacing with underlying hardware, in got a 4x average speed improvement without major accuracy losses.

OpenAI's Rosalind Workbench is now in research preview. We're excited to make the Tamarind Bio catalog of 300+ molecular design models available in Rosalind Workbench, available in ChatGPT from day one.

Introducing the Tamarind Model Router. You want to know which molecular AI model will work best for this specific input, not which model wins on average. Tamarind runs hundreds of scientific AI models, and picks the best tool for the job. buff.ly/lzL1wNY

This week @AnthropicAI showed Claude designing de novo protein binders against 14 of 15 targets. They released the protocol, so we built it into a hosted product you can run today. Introducing the Tamarind Protein Design Agent.

AlphaFold3 is no longer the best model. Over the past year, we've seen many (mostly companies) claim that their newly released models matched or outperformed AlphaFold3 in some important structure prediction task. A substantive amount of these also underperformed their self-professed benchmarks under independent scrutiny. This is why I was initially skeptical on the results from the ESMFold2, Protenix-v2, and OpenDDE teams' respective releases. Within a few weeks of each others, these groups each released benchmarks showing meaningful outperformance over AF3.

Thrilled to partner with @NVIDIAHealth to bring optimized AI tools for molecular design to more scientists We're making NVIDIA NIM microservices and BioNeMo Agent Toolkit available on @TamarindBio. This means NVIDIA's optimized models are available, alongside the rest of your open, proprietary and fine-tuned molecular design tools.

Big news today: New Boltz models! De novo design, protein binding affinity prediction, ADME prediction We've partnered with the @Boltz_Bio team to make their next generation proprietary models available on @TamarindBio on day 0. Here are the highlights:

We’re proud to share that @TamarindBio has been selected to build, host, and operate the inference infrastructure layer for TuneLab2.0, the next evolution of the platform. @EliLillyandCo TuneLab is a first-of-its-kind, collaborative AI/ML drug discovery platform, bringing models trained on over $1B worth of Lilly proprietary data to the biotech ecosystem. Tamarind will power TuneLab’s scalable drug discovery workflows and model inference.

ESMFold2 and the ESM-C family, now available for use! We’ve partnered with @biohub (the ESM team’s new home), to provide day 1 access to their newly open-sourced series of models. The family of models show best-in-class results for structure prediction, de novo design, and protein-language model tasks.

We've launched 300+ molecular AI models on Tamarind. Today, we're giving you the system we use to do it. Introducing Custom Tools 🧵

A new best-in-class structure predictor AND de novo design protocol Protenix-v2 claims to outperform AlphaFold3 in antibody-antigen structure prediction tasks, showing a 13% increase over its previous generation in DockQ scores. Available on @tamarindbio today. Protenix-v2 with only 5 seeds beats Protenix-v1 with 1000 seeds on antibody–antigen prediction. This implies a technical improvement, while not needing to massively scale inference of a given model like other providers previously showed. In addition, the authors use Protenix-v2 as a scoring and ranking mechanism for de novo antibody design. They report a 100% target-level success rate on the current soluble-target panel, meaning at least one confirmed binder for every tested target, with BLI-confirmed VHH-Fc hit rates from 2% to 48%. They also show that epitope choice matters a lot: on AMBP, one epitope gave 4% hit rate and another 48%. The GPCR result is probably the most impressive experimental result in the paper. With only 16–30 tested designs per target, the protocol shows VHH-Fc hit rates of 16%, 62%, 40%, and 88% across four GPCRs, and corresponding mAb hit rates of 0%, 17%, 50%, and 44%. They also report a best GPRC5D VHH-Fc binder of 112 pM under avidity conditions. Congratulations to the @ai4s_protenix team on the release!

Today we announce the Tamarind Bio assay portal: The wet lab, now driven by software We’ve partnered with @AAlphaBio, @adaptyvbio, and @Ginkgo to bring protein and antibody assays directly into Tamarind, making it much easier to move from computational design to real experimental feedback. Protein design is not bottlenecked by generating candidates, but by validating them quickly enough to learn from them. We’re starting with the workhorse experiments: protein-protein binding affinity characterization, developability, expression, and stability. The Assay Portal helps scientists: Get fast, low-friction, cost-effective validation of designed proteins and antibodies, transparent pricing without needing separate MSAs Specialize models on their own experimental data for affinity maturation, developability, and property optimization Run lab-in-the-loop campaigns where each assay result improves the next design cycle Turn wet lab data into model training infrastructure, including RL environments and large-scale datasets for pretraining As computational molecular design matures, we believe integration between wet lab feedback and continuous learning will yield the highest quality results. That’s why we’re excited to bring the unique, differentiated capabilities of our partners to the leading biopharma R&D organizations.

Molecular design for AI agents: announcing the Tamarind MCP Server. Today, scientists can use the @tamarindbio library of 250+ molecular design tools(Boltz, AlphaFold, RFdiffusion...) in any AI chat interface. We serve not just open-source models, but the internal protocols your team has onboarded to Tamarind. Any tool added to Tamarind is then available across the MCP server, Tamarind web app, and API, so it can be used in chat-based agents, multi-step pipelines, ELNs, and LIMS-connected workflows. Our goal is simple: make Tamarind the place where scientists can access the BioAI tools they need, wherever they want to work, while we handle the infrastructure. Many users have already incorporated our MCP into their internal AI agents, along with community efforts like Blatant-Why building apps on top of Tamarind. Try out our tooling for antibody design, small molecule virtual screening, developability/ADMET scoring and more!

Grand Challenges in Computational Small Molecule Drug Discovery This work, a massive undertaking two years in the making, surveys scientific and technical problems where better prediction would materially improve drug discovery outcomes. Benchmarks of methods or models are certainly useful, but we've still not agreed which problem spaces AI can be applied to actually mean "better" drug discovery. Chemistry: synthesis planning, process chemistry, covalent inhibitor design, chemical stability/degradation Structure: crystal packing/polymorphism, protein structure, protein dynamics, protein–ligand pose prediction, cryptic pocket discovery Energy: binding affinity, selectivity, kinetics, allostery/agonism Pharmacology: pKa, solubility/aggregation/permeability, plasma protein binding and volume of distribution, clearance, oral bioavailability, metabolism, toxicity, dose prediction, PK/PD The authors propose that the AI-led transformation will come from solving specific, measurable problems as opposed to fully end-to-end black box solutions. Even in the world where the latter comes true, these challenges are highly valuable evaluations for the efficacy of future protocols. For each challenge the authors outline: • the underlying physical problem • why it matters • the current state of the field • inputs, outputs, and data types • metrics that would define meaningful progress Congratulations on the preprint Woody Sherman, Connor W. Coley, and co-authors. ----- @tamarindbio is a collection of 250+ molecular design tools such as AlphaFold and most of current best solutions to the grand challenges discussed in the paper, accessible via web interface and API.

AlphaFold3 performance, are we there yet? Protenix-v1, IntelliFold, and more claim AF3-level accuracy on a diverse tasks. While they've shown some meaningful improvements, our findings (especially AbAg complexes) show they aren't fully there yet. Let's look at the benchmarks! OpenFold3 preview 2 is out today, available on @tamarindbio along with every other protocol mentioned here.