Blitz: Accelerating the Diffusion Modules of AlphaFold3 and Boltz-2

Vinay Saji Mathew; Soundar Kumara; Gretta Kellogg; William K. M. Lai

cs.LGOct 16, 2026

Blitz: Accelerating the Diffusion Modules of AlphaFold3 and Boltz-2

Authors: Vinay Saji Mathew, Soundar Kumara, Gretta Kellogg, William K. M. Lai

Organizations: Department of Industrial and Manufacturing Engineering, The Pennsylvania State University · Institute for Computational and Data Sciences, The Pennsylvania State University · Department of Biochemistry, University at Buffalo, State University of New York

Abstract

Diffusion-based protein structure prediction models produce all-atom structures in about 200 denoising steps, and that cost has driven a number of few-step samplers. We present Blitz, a few-step distillation framework that uses a geometric noise schedule to distribute steps more evenly across noise scales, then learns those updates through multistep consistency training. Physical steering and target-free candidate selection help reduce the remaining geometric defects at inference. We apply Blitz to AlphaFold 3 and Boltz-2, reducing denoising from 200 steps to eight. Because published few-step results use different benchmarks and scoring procedures, we compare both distilled models with their teachers and other few-step methods under a common evaluation protocol covering structural accuracy, interface quality, ligand placement, and stereochemical validity. Blitz retains teacher-level structural accuracy, with AF3-Blitz outperforming all evaluated few-step baselines and exceeding its teacher on several accuracy metrics.

Figures & tables

Figure 1: Selected 8V9Q example ( Collette et al., 2024 ) : experimental reference, (a) released DeCAF, and (b) its Boltz-1x teacher. Scores show lDDT without $\to$ with stereochemical checks.

		PXM $\uparrow$	OST lDDT $\uparrow$		Events $\downarrow$		PB-valid $\uparrow$
Sampler	NFE	lDDT	Unchecked	Checked	/target	lDDT-PLI $\uparrow$	valid
AF3 teacher	200	0.8554	0.8626	0.8615	0.7	0.6446	44.8%
AF3-Blitz [8]	8	0.8579	0.8653	0.8635	4.8	0.6430	42.7%
Protenix	200	0.8512	0.8593	0.8557	4.3	0.6259	45.8%
Protenix one-step ODE	1	0.8503	0.8591	0.5426	1563.6	0.5774	26.0%
Boltz-2x teacher	600	0.8426	0.8514	0.8499	4.4	0.5596	60.9%

Table 1: Accuracy with and without stereochemical checks. Bold highlights geometry failures. NFE counts steering particles per prediction. Evaluation details are in Appendix B .

Figure 2: Allocation matters most at low budgets. The same untrained AF3 denoiser is evaluated with the default and geometric schedules at matched step budgets and positive endpoints.

Figure 3: Local approximation error and final molecular quality. A: Euler error relative to a refined Heun calculation. B–C: final effects of replacing one refined interval with an Euler step. Details are in Appendix I .

	AlphaFold3					Boltz-2
Metric	Teacher	$\rho{=}7$	Blitz [4]	Blitz [8]	Blitz [16]	Teacher	Blitz [8]	Blitz [16]
NFE	200	8	4	8	16	600	8	16
Complex lDDT $\uparrow$	0.8615	0.8603	0.8559	0.8635	0.8634	0.8499	0.8492	0.8533
Backbone lDDT $\uparrow$	0.9136	0.9151	0.9128	0.9150	0.9151	0.9043	0.9055	0.9069
TM-score $\uparrow$	0.8898	0.8884	0.8889	0.8888	0.8894	0.8753	0.8768	0.8777
DockQ $\uparrow$	0.4198	0.4133	0.4185	0.4207	0.4167	0.3929	0.3897	0.3932

Table 2: Accuracy and geometry across sampling budgets.

Figure 4: The same 8V9Q example as Figure 1 : reference, (a) AF3-Blitz, and (b) its AF3 teacher. Scores show unchecked $\to$ stereochemically checked lDDT.

Metric	Boltz-1x	DeCAF default	DeCAF SDE	DeCAF FK	Protenix	Protenix ODE	Protenix Mini
NFE	600	10	10	40	200	1	1
Complex lDDT $\uparrow$	0.8051	0.6741	0.6607	0.6654	0.8533	0.5462	0.6571
Backbone lDDT $\uparrow$	0.8928	0.8869	0.8784	0.8791	0.9071	0.9039	0.8938
TM-score $\uparrow$	0.8708	0.8622	0.8581	0.8577	0.8807	0.8780	0.8680
DockQ $\uparrow$	0.3868	0.3702	0.3626	0.3636	0.4077	0.4026	0.3684
QS-global $\uparrow$	0.7332	0.7247	0.7196	0.7257	0.7561	0.7483	0.7183

Table 3: Other sampler configurations. Metrics match Table 2 .

Figure 5: Similar averages, different failure patterns. A: paired lDDT changes (geometric minus

\rho=7

); zero denotes equal scores. B–C: defect frequency and violation-count tails on matched Boltz-2 benchmark targets. A uses a signed-log change axis; C uses log-scaled counts and percentages, with a linear region near zero count. Zero counted violations does not imply full chemical validity.

Appendix figures & tables17 assets

Supplementary material from the paper’s appendix.

Appendix

Parameter	Setting
Trainable parameters	Diffusion module only
Initialization	Pretrained teacher weights
Student segments	8
Teacher-grid refinement	200 to 350 edges, log-linear
Optimization budget	3,000 steps
Optimizer	Adam/AdamW, no weight decay

Appendix

Table 4: Eight-segment Blitz training recipe and AF3 sampling minima.

Model	P	P+L	P+NA	P+NA+L
AlphaFold 3 teacher	0.8510	0.8841	0.7686	0.7190
AlphaFold 3 ( $\rho{=}7$ )	0.8506	0.8835	0.7571	0.7040
AlphaFold 3-Blitz [4]	0.8447	0.8820	0.7454	0.6922
AlphaFold 3-Blitz [8]	0.8537	0.8864	0.7647	0.7134
AlphaFold 3-Blitz [16]	0.8544	0.8852	0.7696	0.7165
Boltz-2 teacher	0.8432	0.8698	0.7563	0.7066

Appendix

Table 5: Stereochecked complex lDDT (

\uparrow

) by molecular composition in protein-containing complexes. Boltz-2 benchmark (

\sim

2,300 targets).

Model	P	P+L	P+NA	P+NA+L
AlphaFold 3 teacher	0.8220	0.8760	0.7786	0.7933
AlphaFold 3-Blitz [8]	0.8232	0.8776	0.7844	0.7927
AlphaFold 3-Blitz [16]	0.8232	0.8766	0.7866	0.7923
Boltz-1x teacher	0.7651	0.8144	0.6560	0.6844
DeCAF default	0.6270	0.6934	0.5146	0.5598
DeCAF FK	0.6255	0.6855	0.5109	0.5433

Appendix

Table 5: Stereochecked complex lDDT (

\uparrow

) by molecular composition in protein-containing complexes. RecentPDB384.

	AF3			Boltz	DeCAF		Protenix
Metric	Teacher	Blitz [8]	Blitz [16]	1x	Def.	FK	Def.	Mini
cLDDT $\uparrow$	.8320	.8304	.8345	.8087	.6914	.6955	.8298	.6451
bbLDDT $\uparrow$	.8685	.8722	.8712	.8790	.8712	.8640	.8704	.8538
TM $\uparrow$	.8756	.8830	.8777	.8881	.8850	.8826	.8781	.8635
QS $\uparrow$	.5748	.5755	.5753	.6060	.6058	.6055	.5995	.5617
DockQ $\uparrow$	.1486	.1495	.1484	.1354	.1323	.1303	.1302	.1385

Appendix

Table 6: Runs N’ Poses: structural accuracy, ligand placement, and geometry.

	AF3			Boltz	DeCAF			Protenix
Metric	Teacher	Blitz [8]	Blitz [16]	1x	Def.	FK	SDE	Def.	ODE	Mini
cLDDT $\uparrow$	.8286	.8273	.8286	.7296	.6051	.5993	.5954	.8231	.4680	.5979
bbLDDT $\uparrow$	.8948	.8961	.8958	.8673	.8637	.8573	.8573	.8866	.8822	.8683
TM $\uparrow$	.8572	.8590	.8567	.8287	.8240	.8191	.8226	.8500	.8486	.8277
QS $\uparrow$	.7482	.7502	.7466	.6986	.6922	.6959	.6891	.7257	.7167	.6901
DockQ $\uparrow$	.3428	.3431	.3395	.3036	.2969	.2936	.2957	.3258	.3275	.2986

Appendix

Table 7: RecentPDB384: structural accuracy, ligand placement, and geometry.

Figure 6: A small subset carries much of the clash burden. Left: cumulative clashes ranked by target. Middle: counts for 8oq5. Right: means with all targets and after omitting 8oq5 as an influence diagnostic. Every primary result retains this target.

Figure 7: Severity and incidence answer different questions. Top: targets with a counted defect. Bottom: mean count among affected targets.

Figure 8: The full violation distribution. Curves show the fraction of matched targets above each count. The value at zero is incidence; no extreme target is removed.

Figure 9: Geometry across target sizes. Mean total count and defect incidence within predefined reference-residue bins, using identical targets for all four methods.

	AlphaFold3					Boltz-2
Metric	Teacher	$\rho{=}7$	Blitz [4]	Blitz [8]	Blitz [16]	Teacher	Blitz [8]	Blitz [16]
NFE	200	8	4	8	16	600	8	16
cLDDT $\uparrow$	0.8615	0.8550	0.8512	0.8615	0.8624	0.8503	0.8476	0.8525
bbLDDT $\uparrow$	0.9136	0.9147	0.9128	0.9144	0.9150	0.9047	0.9061	0.9074
TM $\uparrow$	0.8898	0.8902	0.8898	0.8901	0.8909	0.8758	0.8789	0.8792
DockQ $\uparrow$	0.4198	0.4163	0.4220	0.4216	0.4168	0.3927	0.3929	0.3948

Appendix

Table 8: Native-confidence selection on the Boltz-2 benchmark.

Metric	Teacher	$\rho=7$	Blitz [4]	Blitz [8]	Blitz [16]
cLDDT $\uparrow$	[0.857, 0.866]	[0.855, 0.865]	[0.850, 0.861]	[0.859, 0.868]	[0.859, 0.868]
bbLDDT $\uparrow$	[0.909, 0.918]	[0.911, 0.919]	[0.908, 0.917]	[0.910, 0.919]	[0.911, 0.919]
TM $\uparrow$	[0.882, 0.897]	[0.881, 0.896]	[0.881, 0.896]	[0.881, 0.896]	[0.882, 0.897]
DockQ $\uparrow$	[0.403, 0.437]	[0.397, 0.430]	[0.402, 0.435]	[0.404, 0.438]	[0.400, 0.433]
QS $\uparrow$	[0.765, 0.791]	[0.763, 0.788]	[0.767, 0.794]	[0.763, 0.790]	[0.765, 0.791]
ICS $\uparrow$	[0.410, 0.443]	[0.405, 0.438]	[0.412, 0.446]	[0.410, 0.443]	[0.408, 0.441]

Appendix

Table 9: Uncertainty in mean scores and paired comparisons. A. Mean uncertainty: AF3.

Metric	Teacher	Blitz [8]	Blitz [16]
cLDDT $\uparrow$	[0.845, 0.855]	[0.844, 0.854]	[0.848, 0.858]
bbLDDT $\uparrow$	[0.900, 0.909]	[0.901, 0.910]	[0.902, 0.911]
TM $\uparrow$	[0.867, 0.883]	[0.869, 0.885]	[0.870, 0.885]
DockQ $\uparrow$	[0.376, 0.409]	[0.373, 0.406]	[0.377, 0.410]
QS $\uparrow$	[0.724, 0.751]	[0.721, 0.749]	[0.722, 0.749]
ICS $\uparrow$	[0.372, 0.404]	[0.369, 0.401]	[0.370, 0.402]

Appendix

Table 9: Uncertainty in mean scores and paired comparisons. B. Mean uncertainty: Boltz-2.

Metric	Boltz-1x	DeCAF	DeCAF SDE	DeCAF FK	Protenix	Protenix ODE	Protenix Mini
cLDDT $\uparrow$	[0.799, 0.811]	[0.666, 0.682]	[0.653, 0.668]	[0.658, 0.673]	[0.848, 0.859]	[0.538, 0.554]	[0.651, 0.663]
bbLDDT $\uparrow$	[0.888, 0.897]	[0.882, 0.891]	[0.874, 0.883]	[0.874, 0.884]	[0.902, 0.912]	[0.899, 0.909]	[0.889, 0.899]
TM $\uparrow$	[0.863, 0.878]	[0.854, 0.870]	[0.850, 0.866]	[0.849, 0.866]	[0.873, 0.888]	[0.870, 0.886]	[0.860, 0.876]
DockQ $\uparrow$	[0.372, 0.403]	[0.355, 0.385]	[0.348, 0.377]	[0.349, 0.379]	[0.391, 0.424]	[0.387, 0.419]	[0.353, 0.384]
QS $\uparrow$	[0.719, 0.746]	[0.710, 0.738]	[0.705, 0.733]	[0.712, 0.739]	[0.743, 0.769]	[0.735, 0.762]	[0.704, 0.732]
ICS $\uparrow$	[0.369, 0.400]	[0.359, 0.389]	[0.355, 0.385]	[0.360, 0.391]	[0.391, 0.424]	[0.384, 0.416]	[0.353, 0.384]

Appendix

Table 9: Uncertainty in mean scores and paired comparisons. C. Mean uncertainty: other methods.

Metric	AF3 [4]	AF3 [8]	AF3 [16]	Boltz-2 [8]	Boltz-2 [16]
cLDDT $\uparrow$	-.0056 [-.0071, -.0042]	+.0020 [.0012, .0028]	+.0019 [.0013, .0026]	-.0007 [-.0020, .0005]	+.0034 [.0025, .0044]
bbLDDT $\uparrow$	-.0008 [-.0017, .0002]	+.0014 [.0007, .0021]	+.0015 [.0009, .0021]	+.0012 [.0001, .0022]	+.0025 [.0016, .0034]
TM $\uparrow$	-.0009 [-.0026, .0007]	-.0010 [-.0027, .0005]	-.0004 [-.0020, .0011]	+.0016 [-.0003, .0034]	+.0024 [.0006, .0042]
DockQ $\uparrow$	-.0013 [-.0037, .0012]	+.0009 [-.0017, .0033]	-.0031 [-.0057, -.0006]	-.0031 [-.0060, -.0002]	+.0003 [-.0024, .0030]
QS $\uparrow$	+.0021 [-.0003, .0045]	-.0016 [-.0041, .0009]	.0000 [-.0024, .0023]	-.0027 [-.0059, .0003]	-.0019 [-.0049, .0010]
ICS $\uparrow$	+.0026 [.0003, .0048]	.0000 [-.0023, .0023]	-.0016 [-.0037, .0006]	-.0027 [-.0056, -.0001]	-.0020 [-.0046, .0005]

Appendix

Table 9: Uncertainty in mean scores and paired comparisons. D. Student minus teacher.

Figure 10: Noise grids and minimum-noise sensitivity. A–B: eight positive nodes and seven relative intervals, with maximum noise 2,560 Å and minimum 0.064 Å; terminal zero is separate. C–F: accuracy and geometry at eight evaluations as the minimum noise changes. Bands are target-bootstrap intervals. Bond and angle rates are per 100 parameterized residues; clashes are per 100 atoms.

Figure 11: Local defects and reference sensitivity. Top: refined starting states. Bottom: coarse starting states. Columns show single-step Euler and Heun defects against a 128-subdivision reference, followed by the 64-to-128-subdivision reference difference. Curves average candidates within targets; bands resample targets. The horizontal axis gives interval midpoints in noise.

Figure 12: How one coarse interval affects the final structure. Solid and dashed curves use 200 and 100 subdivisions in the refined calculation. A and C–E show branch minus control; B shows unaligned endpoint atom RMSD. Negative accuracy changes and positive violation changes are worse. Bond and angle rates are per 100 parameterized residues; clashes are per 100 atoms. Comparisons pair targets and candidates; bands resample targets.

Protein Structure Prediction

Protein structure prediction uses computational models to infer three-dimensional protein shapes and complexes from sequences and other biological data.

See all 40 papers

DCFold: Efficient Protein Structure Generation with Single Forward Pass
May 18, 2026Zhe Zhang, Yuanning Feng, Yuxuan Song +3Protein Structure Prediction Protein Structure Generation
Institute for AI Industry Research (AIR), Tsinghua University · Department of Computer Science and Technology, Tsinghua University · School of Computer Science and Technology, Huazhong University of Science and Technology +1
Probing and steering biology across Boltz-1s trunk-diffusion boundary
Aug 11, 2026Piotr Jedryszek, Tongmeng Xie, Adam Winnifrith +5Protein Structure Prediction Activation Steering
Department of Biology, University of Oxford, Oxford, UK · Evolvere Biosciences, London, UK · Independent Researcher +4
Few-step Cofolding with All-Atom Flow Maps
Jun 7, 2026Gianluca Scarpellini, Ron Shprints, Peter Holderrieth +7Protein Structure Prediction Diffusion Distillation
1Genesis Molecular AI · Massachusetts Institute of Technology · Carnegie Mellon University +2

May 18, 2026J/K move · Enter open · S save

DCFold: Efficient Protein Structure Generation with Single Forward Pass

Zhe Zhang, Yuanning Feng, Yuxuan Song, Keyue Qiu+2

Institute for AI Industry Research (AIR), Tsinghua University · Department of Computer Science and Technology, Tsinghua University · School of Computer Science and Technology, Huazhong University of Science and Technology+1

Protein Structure Prediction Protein Structure Generation AI-Assisted Scientific Research One-Step Generative Modeling Generative Modeling Molecular Generation+3

Open paper AskPDF arXiv

AlphaFold3 introduces a diffusion-based architecture that elevates protein structure prediction to all-atom resolution with improved accuracy. This state-of-the-art performance has established AlphaFold3 as a foundation model for diverse generation and design tasks. However, its iterative design substantially increases inference time, limiting practical deployment in downstream settings such as virtual screening and protein design. We propose DCFold, a single-step generative model that attains AlphaFold3-level accuracy. Our Dual Consistency training framework, which incorporates a novel Temporal Geodesic Matching (TGM) scheduler, enables DCFold to achieve a 15x acceleration in inference while maintaining predictive fidelity. We validate its effectiveness across both structure prediction and binder design benchmarks.

Abstract

Protein Structure Prediction

Explore similar work

DCFold: Efficient Protein Structure Generation with Single Forward Pass

Probing and steering biology across Boltz-1s trunk-diffusion boundary

Few-step Cofolding with All-Atom Flow Maps