Apex Labs is the research and safety programme that sits alongside the commercial product. We fund external safety research, provide free access to academic groups, run an open-benchmark programme, and are spinning up a civilisational-scale model series for the problems no commercial customer pays for directly.
Each programme has a brief, an application path, and a written commitment from Apex to support it. The brief is published; the commitment is reviewed quarterly.
Free Research-tier access for any accredited academic group.
Citation-grade reproducibility, the published benchmark cases, and the smaller model variants — at zero cost for non-commercial use. We require attribution and basic responsible-use commitments; nothing more.
Grants for external safety research on Aether.
Grants for academic groups working on AI evaluation, biosecurity refusal-corpus methodology, or red-team techniques for autonomous-lab systems. Reviewed by an external panel; decisions and dissents published openly.
Publish reproducible benchmarks against Aether.
We commit to open benchmarks for every new release: the case, the mesh, the solver settings, the extraction script, and the model checkpoint hash. Academic groups can publish head-to-head comparisons under a non-commercial licence; we provide reasonable compute support.
A model series targeted at civilisational-scale problems.
Climate, pandemic preparedness, fusion, food security. Specialised model variants pretrained on the corpora that matter for each problem class, released to qualified non-commercial users at no cost. Pilot launching with three partner institutions in 2026.
Air-gapped access for export-controlled academic work.
For academic groups doing research under ITAR, EAR, or equivalent export-control regimes. Sovereign deployment templates and a clearance-aware engagement model. Applications reviewed quarterly.
Three- to twelve-month fellowships at Apex for academic researchers.
Compute, mentorship, and a salary stipend for academics working on problems adjacent to Aether. Output is published openly; intellectual property follows the fellow. Two cohorts per year.
The Labs civilisational-scale model series targets six problem classes — work we believe matters at the multi-decade horizon and that commercial revenue doesn't naturally fund. Specialised model variants released to qualified non-commercial users at no cost.
Coupled atmosphere–ocean–cryosphere simulation at survey resolution, regional downscaling, extreme-event attribution. The substrate for IPCC-grade analysis you can run yourself.
Pathogen-evolution modelling, transmission dynamics, intervention simulation, biosecurity-grade refusal corpora. Built with biosecurity advisors; never sold; access reviewed by external panel.
Magnetic and inertial confinement physics — MHD equilibrium, edge plasma transport, divertor heat exhaust, breeding-blanket neutronics. Released to fusion programmes with verified safety posture.
Crop-genome breeding simulation, soil microbiome dynamics, drought-resilience screening, supply-chain modelling under climate stress. The science that makes farming the next century viable.
Catalysts for hydrogen and CO₂ reduction, solid-state battery chemistries, photovoltaic and thermovoltaic absorbers. DFT-grade screening with synthesisability constraints.
Computational psychiatry, biomarker discovery, intervention modelling. A specialist variant trained on de-identified longitudinal data with research-use-only licensing.
Labs sits inside Apex but operates under different incentives. Customer urgency does not bend the safety findings; commercial sensitivity does not bend the published benchmarks.
Benchmarks, methodology, reproduction kits are open. Customer data is not. Where the line falls in between is documented case by case.
External red-teams and safety researchers are paid for findings under our vulnerability-disclosure programme. We do not retaliate against good-faith research.
Labs programmes are funded out of operating revenue and report through a separate governance line. Customer relationships do not influence safety findings.
Labs partners get research-tier access to the same Aether that ships to customers. The architecture below is the same backbone in every engagement — different agents on top, same weights underneath.
A snapshot of the briefs Labs partners are running right now. We name the institution, the principal investigator, the workload and the open question the engagement is set up to answer.
Coupled atmosphere-ocean ML emulator
Open question: “Can a simulation-native model close the climate-emulator scaling gap with reanalysis-grade physics?”
Tokamak MHD equilibrium + divertor
Open question: “Where does the model's predicted q-profile diverge from EFIT-reconstructed equilibria, and why?”
Solid-electrolyte conductivity screen
Open question: “Can DFT + ML potential surface dopant compositions that outperform the literature panel by 0.5σ?”
Antibody developability + humanisation
Open question: “What does the model do that a humanisation expert cannot — and where does the expert still beat it?”
Hypersonic re-entry · TPS thermochemistry
Open question: “How does the model handle non-equilibrium air chemistry past Mach 10, where ground-truth is sparse?”
Long-horizon trajectory stability
Open question: “What loss function gives the cleanest forward-roll past 10⁶ steps without compounding drift?”
Reaction mechanism + transition state search
Open question: “Can the model propose a kinetically plausible mechanism that an experienced chemist would have written?”
Long-horizon agent evaluation
Open question: “How do you score a multi-day autonomous engineering run when ground truth is the resulting artefact?”
A slice of the academic partners running grants, fellowships or civilisational-scale pilots with Labs today. Each engagement is named and reviewed; the funded briefs are public.
Application eligibility, IP terms, review timelines and the operating principles of the Labs programme. If yours isn't here, send it to labs@apexworldlabs.com — same-day reply where we can.
Accredited academic groups, non-commercial research institutes, and 501(c)(3) / equivalent non-profits doing work in the civilisational-scale problem classes. Industry researchers should apply via the commercial track.
Research, publication, education, and reproduction of published benchmarks. Commercial use requires a separate licence. Re-distribution of weights is not permitted; re-distribution of derived results is encouraged.
No. The Labs grant agreement explicitly assigns IP to the funded researcher and their institution. We retain a non-exclusive licence to reproduce the published outputs for our own research and educational use.
Two business days to first response. Two weeks for small grants. Four weeks for material funding, which goes to external review.
Public call for nominations, two-year terms, conflict-of-interest disclosure published. Members are named on the Trust Center. Funding decisions are theirs to make; we publish the dissents alongside the consensus.
Labs programmes run on multi-year commitments. If we need to wind a programme down, we publish six months of notice and complete outstanding obligations. We have never closed a programme to date.
The five most-recent papers from the Labs programme and the broader research team. Architecture, methods, benchmarks, safety.
Send a one-page brief to labs@apexworldlabs.com. Decisions in two weeks; partial funding negotiable.