Open source

Biomolecular AI at a fifth of the cost

TT-Bio runs folding, binder design and protein embeddings on Tenstorrent. One Blackhole Galaxy matches an NVIDIA DGX B200 on throughput, for a fifth of the price.

A Tenstorrent Galaxy unit being installed into a rack
throughput per dollar
of purchase price, against a DGX B200
One
codebase
optimized from silicon to serving
1:1
accuracy parity
measured against every reference

Benchmarks

Throughput, every model, every system

Predictions per hour, measured on the same 512-residue protein, with each stack on its own shipped settings. Then divided by what the hardware costs.

Every number here, and how it was measured

Per-model tables, the exact seconds behind every bar, run conditions, the cost model and every price and power source.

See the full benchmarks →

Coverage

Every model, one codebase

Every model runs on the same stack, from the kernels on the chip up to the serving layer. Speed up one part and every model gains.

Structure prediction

Boltz-2Structure and binding affinity
OpenFold3AlphaFold3-family all-atom
Protenix-v2All-atom complex prediction
ESMFold2Rapid single-sequence folding
OpenDDEMulti-chain protein complexes

Binder design

BoltzGenBinders against any target
RFdiffusion3All-atom design and scaffolds

Embeddings

ESM-CSequence embeddings, to 6B
SaProtStructure-aware embeddings

Accuracy

The same answer as the reference

Every model reproduces its official reference implementation, within that reference's own run-to-run spread.

The bar moves, so we measure it

A single reference run is not a target. The reference itself varies between seeds, so we measure that spread and land inside it.

Checked three ways

Submodule activations by correlation, end-to-end outputs on identical noise, and published benchmarks reproduced.

Published, including the misses

Per-target results, thresholds and reproduction commands are in the repository, along with the legs that do not yet pass.

Why it works

Free your GPUs for the work they are still best at

Inference is the half worth moving

Serving is steady, predictable and large. Move it to Tenstorrent and your GPU fleet is left for training and model development, where it earns the most.

Every model fits on one card

No sharding and no interconnect ceiling. A prediction runs on a single AI Processor, so throughput scales linearly with the number of them.

Bring your own models

Agents port a model to the card in a day or two, on your machine. Your weights never leave you.

No HBM, so you can actually buy it

Blackhole pairs GDDR6 with large on-chip SRAM, so the HBM shortage does not gate delivery.

Try it on real targets, free

JapanFold runs TT-Bio on Tenstorrent Galaxy hardware as a public service. Fold something today, with no account and no card.