PacificCompute

Independent infrastructure for AI safety

PacificCompute

Safety needs its own compute.

We provide open model labs with dedicated compute for safety research and evaluation. Our Singapore cluster helps safeguard the open frontier with launch evaluations, research and ongoing technical support.

Independent safety researcher?Request compute at or below cost ↗

First H200 node rented and live · Philanthropically funded

The evaluation gap

01 / 03
6Independent report found
13Developer documentation
21Not yet reviewed

19 coded records · 21 families awaiting review

One square = one ledger entry. Select to inspect.

Reports across the field, not our campaign history or a safety verdict. Inspect the ledger ↗

01 — Compute powering safer open models

Free & Discounted Compute for Evals & Safety Testing

We work with the open frontier from pre-release testing through open-weight releases. We work to accommodate the right testing infrastructure for every lab individually: from lab-hosted endpoints to lab-controlled tenancy with partitioned compute.

Read more on how we work with the open frontier:

02 — The case for dedicated compute

More models. More work to do.

Open models are critical for research, safety and affordable AI for all, but they come with risks: they can be downloaded, modified and run anywhere. Open model releases should be a moment of serious evaluation, and we exist to empower labs to launch more safely without compromising speed or capability.

Open-model releases, by month

02 / 03
Monthly model-family release counts. A data table follows the chart. Likes are a popularity filter, not a capability measure.010202024-09: 12 / 5Sept 242024-10: 5 / 22024-11: 3 / 22024-12: 5 / 2Dec2025-01: 10 / 42025-02: 2 / 12025-03: 4 / 1Mar2025-04: 11 / 32025-05: 4 / 42025-06: 6 / 0Jun2025-07: 7 / 42025-08: 10 / 32025-09: 14 / 4Sept 252025-10: 7 / 42025-11: 6 / 02025-12: 8 / 2Dec2026-01: 11 / 52026-02: 9 / 52026-03: 5 / 3Mar2026-04: 16 / 72026-05: 5 / 22026-06: 15 / 6Jun2026-07: 7 / 02026-08: 10 / 4Aug
View the numbers
Complete months only
Month250+1,000+
2024-09125
2024-1052
2024-1132
2024-1252
2025-01104
2025-0221
2025-0341
2025-04113
2025-0544
2025-0660
2025-0774
2025-08103
2025-09144
2025-1074
2025-1160
2025-1282
2026-01115
2026-0295
2026-0353
2026-04167
2026-0552
2026-06156
2026-0770
2026-08104
Source: Hugging Face. Model sizes consolidated; product tiers retained. Popularity ≠ capability.Methods

From a question to a finding.

Working with us
01

Propose a Question

A lab, evaluator or researcher can propose a model and a safety question. We’ll help scope the work, access the model and set up the right reporting before work starts.

02

Run the evaluation

Dedicated GPUs sited in Singapore will power your work’s compute needs.

03

Make the findings useful

We’ll help you review your findings, reproduce and validate them, and send them on to the open frontier lab. The lab itself decides how to respond, with no obligation to publish.

03 — A concrete use of funding

Fund safety at the open frontier

A suite of safety evaluations costs $200,000 all-in, including researcher time, compute resources and the follow-up that turns a signal into useful actions.

Interested in funding safety work on both sides of the Pacific? Reach out to scope a grant.

One scoped evaluation

03 / 03
  • Evaluation compute and environmentsIncluded
  • Research and delivery staffIncluded
  • Investigation, reproduction and scoped retestingIncluded
  • Storage and data transferIncluded
  • Allocated operations and contingencyIncluded
  • Private findings and resource statementIncluded
All-in evaluation price$200k
USD. For the agreed campaign scope. A programme price, not an audited average cost or a promise of unlimited testing.Scope ↗

Put compute to work.

Bring a model, an evaluation, or the funding to make it happen.