Galactica vs Terracotta

Five open-weight checkpoints from 125M to 120B parameters give researchers a science-trained language model built from 106 billion curated tokens across 48 million papers, textbooks, encyclopedias, and knowledge bases. Galactica stores, combines, and reasons across modalities including LaTeX, Python code, SMILES formulas, and amino acid sequences inside one decoder-only Transformer architecture. Meta's Papers with Code team open sourced the weights for researchers who want to study how language models organize scientific knowledge.

General-purpose LLMs train on broad web crawls where social chatter can dominate the token budget. Galactica used only curated open-access science sources, which the paper argues lets it train for multiple epochs without overfitting. Meta removed the public web demo three days after launch in November 2022 when critics showed confident but fabricated citations and equations, so today you download checkpoints from Hugging Face rather than chat through a hosted interface.

Machine learning researchers can reproduce benchmark numbers from the arXiv paper, including 77.6% on PubMedQA and 52.9% on MedMCQA dev. Computational biologists and chemists can test protein annotation and molecule tasks through Galactica's specialized tokens for amino sequences and SMILES strings. Graduate students exploring scientific QA, citation prediction, or math word problems can experiment with sizes up to 120 billion parameters without training from scratch.

Terracotta is an Infrastructure as Code governance tool that audits every pull request before merge, checking Terraform and OpenTofu changes against live cloud resources, remote state, and your team's governance policies. It installs as a GitHub or GitLab app, posts findings in the PR thread, and builds a tamper-evident audit trail regulators can export. The product targets platform engineering teams that need security, drift, cost, and compliance checks without rewriting CI pipelines.

Static scanners like Checkov or tfsec lint HCL syntax and known misconfigurations, but they never compare a plan to what is actually running in AWS. Terracotta closes that gap by correlating code, Terraform state, and live resources so drift, cross-PR conflicts, and hidden blast radius show up before anyone clicks merge. Its guardrails are written in plain English rather than Rego or Sentinel, which lowers the bar for teams that lack a dedicated policy-as-code engineer.

DevOps leads and platform engineers at regulated shops use Terracotta to block public S3 buckets, open SSH rules, and unapproved cost spikes at review time instead of in production. Security and compliance teams get a fleet-wide dashboard with drift posture, policy compliance rates, and exportable records for SOC 2 or HIPAA audits. Developers keep working inside GitHub or GitLab because findings arrive as PR comments, not another portal to check.

Beacon, Terracotta's in-PR chat assistant, answers questions about specific findings using context from the repo, plan output, and drift reports. The Platform tier adds unlimited drift repos, IAM and blast-radius analysis, Slack notifications, and a command center dashboard for $49 per engineer per month. Enterprise customers can run Terracotta self-hosted with SSO, SAML, and custom integrations for HCP Terraform or CircleCI.