diff --git a/benchmark/Manifest.toml b/benchmark/Manifest.toml
index fa3aef6e..ad9a0b7b 100644
--- a/benchmark/Manifest.toml
+++ b/benchmark/Manifest.toml
@@ -2,11 +2,11 @@
 
 julia_version = "1.9.1"
 manifest_format = "2.0"
-project_hash = "14fe4b55e0aa680d5c90f646c1a87c8fc8737479"
+project_hash = "0cb0756144aac73ae8e2d06d9a0f6567a7a2f964"
 
 [[deps.ADNLPModels]]
 deps = ["ADTypes", "ForwardDiff", "LinearAlgebra", "NLPModels", "Requires", "ReverseDiff", "SparseArrays", "SparseConnectivityTracer", "SparseMatrixColorings"]
-git-tree-sha1 = "ad4682ad3f6da4246a5a5408593e5824d949e5a0"
+git-tree-sha1 = "2b582670fb51216d8d000c6de72934d1f68c4e7c"
 repo-rev = "main"
 repo-url = "https://github.com/JuliaSmoothOptimizers/ADNLPModels.jl"
 uuid = "54578032-b7ea-4c30-94aa-7cbd1cce6c9a"
diff --git a/benchmark/Project.toml b/benchmark/Project.toml
index d561c1ab..a3e2c200 100644
--- a/benchmark/Project.toml
+++ b/benchmark/Project.toml
@@ -1,6 +1,5 @@
 [deps]
 ADNLPModels = "54578032-b7ea-4c30-94aa-7cbd1cce6c9a"
-BenchmarkProfiles = "ecbce9bc-3e5e-569d-9e29-55181f61f8d0"
 BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
 DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
 Dates = "ade2ca70-3891-5945-98fb-dc099432e06a"
@@ -13,7 +12,6 @@ LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 NLPModels = "a4795742-8479-5a88-8948-cc11e1c8c1a6"
 NLPModelsJuMP = "792afdf1-32c1-5681-94e0-d7bf7a5df49e"
 OptimizationProblems = "5049e819-d29b-5fba-b941-0eee7e64c1c6"
-Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 ReverseDiff = "37e2e3b7-166d-5795-8a7a-e32c996b4267"
 SolverBenchmark = "581a75fa-a23a-52d0-a590-d6201de2218a"
diff --git a/benchmark/benchmark_analyzer/Project.toml b/benchmark/benchmark_analyzer/Project.toml
new file mode 100644
index 00000000..ecbdd021
--- /dev/null
+++ b/benchmark/benchmark_analyzer/Project.toml
@@ -0,0 +1,9 @@
+[deps]
+BenchmarkProfiles = "ecbce9bc-3e5e-569d-9e29-55181f61f8d0"
+BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
+DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
+Dates = "ade2ca70-3891-5945-98fb-dc099432e06a"
+JLD2 = "033835bb-8acc-5ee8-8aae-3f567f8a3819"
+JSON = "682c06a0-de6a-54ab-a142-c8b1cf79cde6"
+SolverBenchmark = "581a75fa-a23a-52d0-a590-d6201de2218a"
+StatsPlots = "f3b207a7-027a-5e70-b257-86293d7955fd"
diff --git a/benchmark/benchmarks.jl b/benchmark/benchmarks.jl
index 90a575f6..225fae26 100644
--- a/benchmark/benchmarks.jl
+++ b/benchmark/benchmarks.jl
@@ -1,13 +1,14 @@
 # Include useful packages
 using ADNLPModels
 using Dates, DelimitedFiles, JLD2, LinearAlgebra, Printf, SparseArrays
-using BenchmarkTools, DataFrames, Plots
+using BenchmarkTools, DataFrames
 #JSO packages
-using NLPModels, BenchmarkProfiles, OptimizationProblems, SolverBenchmark
+using NLPModels, OptimizationProblems, SolverBenchmark
 # Most likely benchmark with JuMP as well
 using JuMP, NLPModelsJuMP
 
 include("problems_sets.jl")
+verbose_subbenchmark = false
 
 # Run locally with `tune!(SUITE)` and then `run(SUITE)`
 const SUITE = BenchmarkGroup()
diff --git a/benchmark/gradient/benchmarks_gradient.jl b/benchmark/gradient/benchmarks_gradient.jl
index e8d907e5..56caf700 100644
--- a/benchmark/gradient/benchmarks_gradient.jl
+++ b/benchmark/gradient/benchmarks_gradient.jl
@@ -51,7 +51,7 @@ for f in benchmark_list
         for pb in problem_sets[s]
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
-          @info " $(pb): $T with $n vars and $m cons"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars and $m cons"
           g = zeros(T, n)
           SUITE["$(fun)"][f][T][s][b][pb] = @benchmarkable $fun(nlp, get_x0(nlp), $g) setup =
             (nlp = set_adnlp($pb, $(name_backend), $(backend), $nscal, $T))
diff --git a/benchmark/hessian/benchmarks_coloring.jl b/benchmark/hessian/benchmarks_coloring.jl
index 3ff43148..de953c0c 100644
--- a/benchmark/hessian/benchmarks_coloring.jl
+++ b/benchmark/hessian/benchmarks_coloring.jl
@@ -49,7 +49,7 @@ for f in benchmark_list
         for pb in problem_sets[s]
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
-          @info " $(pb): $T with $n vars and $m cons"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars and $m cons"
           SUITE["$(fun)"][f][T][s][b][pb] =
             @benchmarkable set_adnlp($pb, $(name_backend), $backend, $nscal, $T)
         end
diff --git a/benchmark/hessian/benchmarks_hessian.jl b/benchmark/hessian/benchmarks_hessian.jl
index 53be8982..dc2368e7 100644
--- a/benchmark/hessian/benchmarks_hessian.jl
+++ b/benchmark/hessian/benchmarks_hessian.jl
@@ -40,7 +40,7 @@ for f in benchmark_list
         for pb in problem_sets[s]
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
-          @info " $(pb): $T with $n vars"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars"
           SUITE["$(fun)"][f][T][s][b][pb] = @benchmarkable $fun(nlp, get_x0(nlp)) setup =
             (nlp = set_adnlp($pb, $(name_backend), $backend, $nscal, $T))
         end
diff --git a/benchmark/hessian/benchmarks_hessian_lagrangian.jl b/benchmark/hessian/benchmarks_hessian_lagrangian.jl
index 7801657c..daa742c6 100644
--- a/benchmark/hessian/benchmarks_hessian_lagrangian.jl
+++ b/benchmark/hessian/benchmarks_hessian_lagrangian.jl
@@ -40,7 +40,7 @@ for f in benchmark_list
         for pb in problem_sets[s]
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
-          @info " $(pb): $T with $n vars and $m cons"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars and $m cons"
           y = 10 * T[-(-1.0)^i for i = 1:m]
           SUITE["$(fun)"][f][T][s][b][pb] = @benchmarkable $fun(nlp, get_x0(nlp), $y) setup =
             (nlp = set_adnlp($pb, $(name_backend), $backend, $nscal, $T))
diff --git a/benchmark/hessian/benchmarks_hessian_residual.jl b/benchmark/hessian/benchmarks_hessian_residual.jl
index 9fc8467a..7058ad50 100644
--- a/benchmark/hessian/benchmarks_hessian_residual.jl
+++ b/benchmark/hessian/benchmarks_hessian_residual.jl
@@ -41,7 +41,7 @@ for f in benchmark_list
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
           nequ = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nls_nequ(n = $(nscal))"))
-          @info " $(pb): $T with $n vars, $nequ residuals and $m cons"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars, $nequ residuals and $m cons"
           v = 10 * T[-(-1.0)^i for i = 1:nequ]
           SUITE["$(fun)"][f][T][s][b][pb] = @benchmarkable $fun(nls, get_x0(nls), $v) setup =
             (nls = set_adnls($pb, $(name_backend), $backend, $nscal, $T))
diff --git a/benchmark/hessian/benchmarks_hprod.jl b/benchmark/hessian/benchmarks_hprod.jl
index 99f7cac9..38049ed2 100644
--- a/benchmark/hessian/benchmarks_hprod.jl
+++ b/benchmark/hessian/benchmarks_hprod.jl
@@ -38,7 +38,7 @@ for f in benchmark_list
         for pb in problem_sets[s]
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
-          @info " $(pb): $T with $n vars"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars"
           v = [sin(T(i) / 10) for i = 1:n]
           Hv = Vector{T}(undef, n)
           SUITE["$(fun)"][f][T][s][b][pb] = @benchmarkable $fun(nlp, get_x0(nlp), $v, $Hv) setup =
diff --git a/benchmark/hessian/benchmarks_hprod_lagrangian.jl b/benchmark/hessian/benchmarks_hprod_lagrangian.jl
index daee07f1..3f9dd730 100644
--- a/benchmark/hessian/benchmarks_hprod_lagrangian.jl
+++ b/benchmark/hessian/benchmarks_hprod_lagrangian.jl
@@ -38,7 +38,7 @@ for f in benchmark_list
         for pb in problem_sets[s]
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
-          @info " $(pb): $T with $n vars"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars"
           y = 10 * T[-(-1.0)^i for i = 1:m]
           v = [sin(T(i) / 10) for i = 1:n]
           Hv = Vector{T}(undef, n)
diff --git a/benchmark/jacobian/benchmarks_coloring.jl b/benchmark/jacobian/benchmarks_coloring.jl
index 69974255..bf0e7023 100644
--- a/benchmark/jacobian/benchmarks_coloring.jl
+++ b/benchmark/jacobian/benchmarks_coloring.jl
@@ -49,7 +49,7 @@ for f in benchmark_list
         for pb in problem_sets[s]
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
-          @info " $(pb): $T with $n vars and $m cons"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars and $m cons"
           SUITE["$(fun)"][f][T][s][b][pb] =
             @benchmarkable set_adnlp($pb, $(name_backend), $backend, $nscal, $T)
         end
diff --git a/benchmark/jacobian/benchmarks_jacobian.jl b/benchmark/jacobian/benchmarks_jacobian.jl
index 16f9b5a1..402653a0 100644
--- a/benchmark/jacobian/benchmarks_jacobian.jl
+++ b/benchmark/jacobian/benchmarks_jacobian.jl
@@ -36,7 +36,7 @@ for f in benchmark_list
         for pb in problem_sets[s]
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
-          @info " $(pb): $T with $n vars and $m cons"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars and $m cons"
           SUITE["$(fun)"][f][T][s][b][pb] = @benchmarkable $fun(nlp, get_x0(nlp)) setup =
             (nlp = set_adnlp($pb, $(name_backend), $backend, $nscal, $T))
         end
diff --git a/benchmark/jacobian/benchmarks_jacobian_residual.jl b/benchmark/jacobian/benchmarks_jacobian_residual.jl
index 24e11e02..a2fb6f89 100644
--- a/benchmark/jacobian/benchmarks_jacobian_residual.jl
+++ b/benchmark/jacobian/benchmarks_jacobian_residual.jl
@@ -37,7 +37,7 @@ for f in benchmark_list
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
           nequ = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nls_nequ(n = $(nscal))"))
-          @info " $(pb): $T with $n vars, $nequ residuals and $m cons"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars, $nequ residuals and $m cons"
           SUITE["$(fun)"][f][T][s][b][pb] = @benchmarkable $fun(nls, get_x0(nls)) setup =
             (nls = set_adnls($pb, $(name_backend), $backend, $nscal, $T))
         end
diff --git a/benchmark/jacobian/benchmarks_jprod.jl b/benchmark/jacobian/benchmarks_jprod.jl
index b5b6e315..08f20b56 100644
--- a/benchmark/jacobian/benchmarks_jprod.jl
+++ b/benchmark/jacobian/benchmarks_jprod.jl
@@ -38,7 +38,7 @@ for f in benchmark_list
         for pb in problem_sets[s]
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
-          @info " $(pb): $T with $n vars and $m cons"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars and $m cons"
           Jv = Vector{T}(undef, m)
           v = 10 * T[-(-1.0)^i for i = 1:n]
           SUITE["$(fun)"][f][T][s][b][pb] = @benchmarkable $fun(nlp, get_x0(nlp), $v, $Jv) setup =
diff --git a/benchmark/jacobian/benchmarks_jprod_residual.jl b/benchmark/jacobian/benchmarks_jprod_residual.jl
index 810fc400..77208055 100644
--- a/benchmark/jacobian/benchmarks_jprod_residual.jl
+++ b/benchmark/jacobian/benchmarks_jprod_residual.jl
@@ -39,7 +39,7 @@ for f in benchmark_list
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
           nequ = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nls_nequ(n = $(nscal))"))
-          @info " $(pb): $T with $n vars, $nequ residuals and $m cons"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars, $nequ residuals and $m cons"
           Jv = Vector{T}(undef, nequ)
           v = 10 * T[-(-1.0)^i for i = 1:n]
           SUITE["$(fun)"][f][T][s][b][pb] = @benchmarkable $fun(nlp, get_x0(nlp), $v, $Jv) setup =
diff --git a/benchmark/jacobian/benchmarks_jtprod.jl b/benchmark/jacobian/benchmarks_jtprod.jl
index aa4eb9b0..927b3eb0 100644
--- a/benchmark/jacobian/benchmarks_jtprod.jl
+++ b/benchmark/jacobian/benchmarks_jtprod.jl
@@ -38,7 +38,7 @@ for f in benchmark_list
         for pb in problem_sets[s]
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
-          @info " $(pb): $T with $n vars and $m cons"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars and $m cons"
           Jtv = Vector{T}(undef, n)
           v = 10 * T[-(-1.0)^i for i = 1:m]
           SUITE["$(fun)"][f][T][s][b][pb] = @benchmarkable $fun(nlp, get_x0(nlp), $v, $Jtv) setup =
diff --git a/benchmark/jacobian/benchmarks_jtprod_residual.jl b/benchmark/jacobian/benchmarks_jtprod_residual.jl
index aa6e64a4..144455bd 100644
--- a/benchmark/jacobian/benchmarks_jtprod_residual.jl
+++ b/benchmark/jacobian/benchmarks_jtprod_residual.jl
@@ -39,7 +39,7 @@ for f in benchmark_list
           n = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nvar(n = $(nscal))"))
           m = eval(Meta.parse("OptimizationProblems.get_" * pb * "_ncon(n = $(nscal))"))
           nequ = eval(Meta.parse("OptimizationProblems.get_" * pb * "_nls_nequ(n = $(nscal))"))
-          @info " $(pb): $T with $n vars, $nequ residuals and $m cons"
+          verbose_subbenchmark && @info " $(pb): $T with $n vars, $nequ residuals and $m cons"
           Jtv = Vector{T}(undef, n)
           v = 10 * T[-(-1.0)^i for i = 1:nequ]
           SUITE["$(fun)"][f][T][s][b][pb] = @benchmarkable $fun(nlp, get_x0(nlp), $v, $Jtv) setup =
diff --git a/benchmark/run_analyzer.jl b/benchmark/run_analyzer.jl
new file mode 100644
index 00000000..adbe5328
--- /dev/null
+++ b/benchmark/run_analyzer.jl
@@ -0,0 +1,65 @@
+using Pkg
+Pkg.activate("benchmark/benchmark_analyzer")
+Pkg.instantiate()
+using BenchmarkTools, Dates, JLD2, JSON, Plots, StatsPlots
+
+# name of the result file:
+name = ""
+resultpath = joinpath(dirname(@__FILE__), "results")
+if name == ""
+  name = replace(readdir(resultpath)[end], ".jld2" => "", ".json" => "")
+end
+
+@load joinpath(dirname(@__FILE__), "results", "$name.jld2") result
+t = BenchmarkTools.load(joinpath(dirname(@__FILE__), "results", "$name.json"))
+
+# plots
+using StatsPlots
+plot(t) # ou can use all the keyword arguments from Plots.jl, for instance st=:box or yaxis=:log10.
+
+@info "Available benchmarks"
+df_results = Dict{String, Dict{Symbol, DataFrame}}()
+for benchmark in keys(result)
+  result_bench = result[benchmark] # one NLPModel API function
+  for benchmark_list in keys(result_bench)
+    for type_bench in keys(result_bench[benchmark_list])
+      for set_bench in keys(result_bench[benchmark_list][type_bench])
+        @info "$benchmark/$benchmark_list for type $type_bench on problem set $(set_bench)"
+        bench = result_bench[benchmark_list][type_bench][set_bench]
+        df_results["$(benchmark)_$(benchmark_list)_$(type_bench)_$(set_bench)"] = bg_to_df(bench)
+      end
+    end
+  end
+end
+
+function bg_to_df(bench::BenchmarkGroup)
+  solvers = collect(keys(bench)) # "jump", ...
+  nsolvers = length(solvers)
+  problems = collect(keys(bench[solvers[1]]))
+  nprob = length(problems)
+  dfT = Dict{Symbol, DataFrame}()
+  for solver in solvers
+    dfT[Symbol(solver)] = DataFrame(
+      [
+        [median(bench[solver][pb]).time for pb in problems],
+        [median(bench[solver][pb]).memory for pb in problems],
+      ],
+    [:median_time, :median_memory]
+    )
+  end
+  return dfT
+end
+
+using SolverBenchmark, BenchmarkProfiles
+
+# b::BenchmarkProfiles.AbstractBackend = PlotsBackend()
+costs =[
+  df -> df.median_time,
+  df -> df.median_memory,
+]
+costnames = ["median time", "median memory"]
+for key_benchmark in keys(df_results)
+  stats = df_results[key_benchmark]
+  p = profile_solvers(stats, costs, costnames)
+  savefig(p, "$(name)_$(key_benchmark).png")
+end
diff --git a/benchmark/run_local.jl b/benchmark/run_local.jl
index f700ba1e..b62a7404 100644
--- a/benchmark/run_local.jl
+++ b/benchmark/run_local.jl
@@ -1,24 +1,46 @@
-using Pkg, Logging, JLD2, Dates
+using Pkg
 Pkg.activate("benchmark")
-# instantiate
-# up ADNLPModels
+Pkg.instantiate()
+Pkg.update("ADNLPModels")
+using Logging, JLD2, Dates
 
+path = dirname(@__FILE__)
+skip_tune = true
+
+@info "INITIALIZE"
 include("benchmarks.jl")
 
+list_of_benchmark = keys(SUITE)
+# gradient: SUITE[@tagged "grad!"]
+# Coloring benchmark: SUITE[@tagged "hessian_backend" || "hessian_residual_backend" || "jacobian_backend" || "jacobian_residual_backend"]
+# Matrix benchmark: SUITE[@tagged "hessian_backend" || "hessian_residual_backend" || "jacobian_backend" || "jacobian_residual_backend" || "hess_coord!" || "hess_coord_residual!" || "jac_coord!" || "jac_coord_residual!"]
+# Matrix-vector products: SUITE[@tagged "hprod!" || "hprod_residual!" || "jprod!" || "jprod_residual!" || "jtprod!" || "jtprod_residual!"]
+
+for benchmark_in_suite in list_of_benchmark
+  @info "$(benchmark_in_suite)"
+end
+
 @info "TUNE"
-@time with_logger(ConsoleLogger(Error)) do # remove warnings
-  tune!(SUITE)
+if !skip_tune
+  @time with_logger(ConsoleLogger(Error)) do
+    tune!(SUITE)
+    BenchmarkTools.save("params.json", params(suite));
+  end
+else
+  @info "Skip tuning"
+  # https://juliaci.github.io/BenchmarkTools.jl/dev/manual/
+  BenchmarkTools.DEFAULT_PARAMETERS.evals = 1
 end
 
 @info "RUN"
 @time result = with_logger(ConsoleLogger(Error)) do # remove warnings
-  run(SUITE)
+  if "params.json" in (path == "" ? readdir() : readdir(path))
+    loadparams!(suite, BenchmarkTools.load("params.json")[1], :evals, :samples);
+  end
+  run(SUITE, verbose = true)
 end
 
 @info "SAVE BENCHMARK RESULT"
 name = "$(today())_adnlpmodels_benchmark"
 @save "$name.jld2" result
-
-@info "ANALYZE"
-# save the result in a jld2 file?
-# plots
+BenchmarkTools.save("$name.json", result)