LANGUAGE MODEL Google/DeepMind

AlphaFold 2

Protein Structure Prediction Solved

DeepMind's protein structure prediction system that, in late 2020, solved a 50-year-old grand challenge in biology: given a protein's amino acid sequence, predict its 3D folded shape. The 2021 paper + open-source release made these predictions free for any researcher worldwide. Within 3 years, the AlphaFold Database contained predicted structures for nearly all 200 million known proteins.

Cost
Free
Open weights — self-host
How are Intelligence, Speed & Cost bucketed?
Intelligence and Speed buckets are percentile ranks on Artificial Analysis. Cost buckets are fixed dollar thresholds keyed off output-token price ($/M out).
Intelligence
  • Top 1%≤ 1%
  • Top 5%≤ 5%
  • Top 10%≤ 10%
  • Good≤ 25%
  • Medium≤ 50%
  • Below avg> 50%
Speed
  • Top 1%≥ 345 tok/s
  • Top 5%≥ 237 tok/s
  • Top 10%≥ 196 tok/s
  • Good≥ 146 tok/s
  • Medium≥ 90 tok/s
  • Slow< 90 tok/s
Cost
  • Freeopen weights · self-host
  • Low< $1 / M out
  • Moderate$1–5 / M out
  • High≥ $5 / M out

Why it matters

AlphaFold proved that AI could not just match but transform a foundational scientific discipline. The Nobel Prize award in 2024 made the recognition official: this is a Nobel-class contribution to science, not just to AI.

Core Capabilities

Science
Built for biology, chemistry, materials, weather, or math research.
Generative
Produces images, video, audio, or other media.

Context Window

Context window not disclosed.

Availability

API
Not available
Product / App
Not available
Open Source
Released
Enterprise

Pricing Model

Free / self-host
Open weights — pay only for compute.
Self-host

Capability / Performance

Where this model sits relative to the middle 60% of models in the tree. All scores are 0–10 (higher is better).

Lower 20% Upper 80% This model
Lower 20% 20th percentile — 20% of models score below this This model Where the current model lands Upper 80% 80th percentile — only 20% of models score above this Percentile boundaries are computed across every model in the tree that reports the underlying benchmark for each capability.

What it feels like

  • Solved a 50-year grand challenge in biology — protein structure prediction at experimental accuracy
  • CASP14 median GDT score 92.4 vs prior best ~62 — the kind of step-change that ends a research subfield
  • Predicted 200M+ protein structures in AlphaFold DB by 2022 — 1000x more than all experimental data combined
  • Demis Hassabis and John Jumper won the 2024 Nobel Prize in Chemistry — first AI work to do so
  • Open-source code + structure database used by 2M+ researchers worldwide
  • AlphaFold 3 (2024) extended this to protein-ligand and protein-DNA complexes

Best use cases

  • Drug discovery and target validation pipelines
  • Structural biology research where experimental crystallography is too slow / expensive
  • Vaccine and antibody design starting from sequence
  • Teaching how transformer-style attention generalises beyond language

Tools to try

Not ideal for

  • Membrane proteins and intrinsically disordered regions (still imperfect)
  • Direct use as a chemistry / docking model (AlphaFold 3 is more relevant for that)

Model Evolution

View full evolution tree →

Jumper, J. · Evans, R. · Pritzel, A. · Green, T. · et al.