OpenTofu 1.13: Fewer Guesswork Plans, and an IaC Fork That Came of Age
If you have written Terraform or OpenTofu configs for any length of time, you know the exact moment a plan turns from promise into a shrug. You feed the output of one resource into the count or for_each of another, and the CLI answers: "This value is not known until apply." The plan sits there half-finished, full of question marks. Nothing to do but wait.
OpenTofu 1.13, out in late September 2026, takes a real swing at that problem. And it is the crispest reminder in a while that this project is no longer just "the other Terraform"—it is running its own agenda.
Where the fork came from
A quick recap for anyone who joined late. When HashiCorp said in August 2023 that Terraform would move off the Mozilla Public License to its Business Source License, the community did not take it lying down. A fork called OpenTF appeared within days, backed by more than a hundred companies and hundreds of individuals pledging time and money. It relaunched as OpenTofu under the Linux Foundation that September, shipped its first stable release in January 2024, and was accepted into the CNCF in April 2025. The surprising part is how quickly "keeping Terraform open" turned into "building a better Terraform."
The new trick: fewer unknowns at plan time
The headliner in 1.13 is a set of functions designed to shrink the "unknown" zone during planning. OpenTofu has always been conservative: if it cannot prove something is true, it treats it as unknown. The cost is that modules become opaque—one deploy's count stays unresolved until apply, and automation built on top of the plan graph gets downgraded to guesswork.
The new functions, including convert and several assume... forms, let a module author state facts the tool genuinely cannot infer on its own, the way you might tell a colleague "trust me, this is a list of exactly three items." Give the planner enough hints and it can resolve how many instances to create, or which for_each keys exist, without blocking on apply. The team expects the feature to stabilize in 1.14. It is a niche fix on the surface and a quietly big one underneath—fewer unknowable plans means CI and approvals that actually look at the whole picture.
Signs of an independent roadmap
The release also shows the project maturing on its own terms. There is early, experimental work toward built-in linting that the maintainers hope to fold into broader policy checks. A "Symbol Library" experiment explores reusable bundles of values, functions, and type aliases you can import like a module. There are Windows-on-ARM64 builds, a phase-out of 32-bit packages, and WinRM support removed from provisioners—housekeeping that suggests a team no longer mirroring HashiCorp's cadence.
Most tellingly, OpenTofu shortened its release cycle to line up with Go's, so each series enjoys a longer security-support window; the 1.13 line is supported into August 2027. That is the quiet answer to the second-guessing that once followed every fork: Will someone still maintain this in three years? Yes—with a schedule to prove it.
Adopting the tool that used to be Terraform is no longer a political statement. It is just a technical decision, and 1.13 makes that decision easier.