docs: the published benchmarks are re-measured on specsolve 0.6.0

CHANGELOG.md

@@ -8,6 +8,7 @@ it releases that version ([RELEASING.md](https://github.com/fluxopt/specsolve/bl
 ## Upcoming version
 
 - fix(deps): specsolve installs a polars older than 2.0, on which a row's dual can come back empty and a model can take nine times the memory ([#1852](https://github.com/fluxopt/specsolve/pull/1852))
+- docs: the published benchmarks are re-measured on specsolve 0.6.0 ([#1850](https://github.com/fluxopt/specsolve/pull/1850))
 
 ## 0.6.0 (2026-10-05)
 

bench/reproduce.py

@@ -1,7 +1,8 @@
 # /// script
 # requires-python = ">=3.12"
 # dependencies = [
-#   "lpspec[gurobi,linopy] @ git+https://github.com/fluxopt/lpspec@8d27e88b79",
+#   "specsolve[gurobi] @ git+https://github.com/fluxopt/specsolve@aa90239393",
+#   "linopy @ git+https://github.com/PyPSA/linopy@master",
 #   "pyomo>=6.7",
 #   "pytest==9.1.1",
 #   "pytest-benchmem>=0.5",
@@ -17,9 +18,6 @@
 `--locked` refuses to run if the resolution has drifted. The selection is read
 from the `pixi run ladder` task in `pyproject.toml`, and the harness in `bench/`
 runs it, so this needs the repository checked out.
-
-The published numbers predate this file: they were taken against
-`lpspec 0.0.1a61.dev3+gf319cd10f` and a linopy built from a branch.
 """
 
 from __future__ import annotations

bench/results/latest-gurobi-dispatch.ceilings.json

@@ -9,6 +9,6 @@
   "budget": 30.0,
   "memory_budget": 16.0,
   "stopped_by": "memory",
-  "reason": "pyomo took 2.14 GB on dispatch/m, so the next rung projects to 42.9 GB \u2014 over the 16 GB budget"
+  "reason": "pyomo took 2.17 GB on dispatch/m, so the next rung projects to 43.3 GB \u2014 over the 16 GB budget"
  }
 ]