diff --git a/src/NLPModels.jl b/src/NLPModels.jl
index 17299f9b..619b0cda 100644
--- a/src/NLPModels.jl
+++ b/src/NLPModels.jl
@@ -7,7 +7,7 @@ using FastClosures
 # JSO
 using LinearOperators
 
-export AbstractNLPModel, AbstractNLSModel
+export AbstractNLPModel, AbstractNLSModel, AbstractDenseNLPModel
 
 # For documentation purpose
 const OBJECTIVE_HESSIAN = raw"""
@@ -30,6 +30,13 @@ Base type for an optimization model.
 """
 abstract type AbstractNLPModel{T, S} end
 
+"""
+    AbstractDenseNLPModel <: AbstractNLPModel
+
+Base type for a dense optimization model (Jacobian/Hessian stored as dense matrices).
+"""
+abstract type AbstractDenseNLPModel{T, S} <: AbstractNLPModel{T, S} end
+
 """
     AbstractNLSModel <: AbstractNLPModel
 
diff --git a/src/nlp/api.jl b/src/nlp/api.jl
index 22863b45..69f0ac2c 100644
--- a/src/nlp/api.jl
+++ b/src/nlp/api.jl
@@ -1387,7 +1387,9 @@ This is typically used to normalize variables for numerical stability in solvers
 By default, the scaling is model-dependent. If not overridden by the model, a vector of ones 
 is returned. Inspired by the AMPL scaling conventions.
 """
-function varscale end
+function varscale(model::AbstractNLPModel{T, S}) where {T, S}
+  return ones(T, model.meta.nvar)
+end
 
 """
     lagscale(model::AbstractNLPModel)
@@ -1395,7 +1397,9 @@ function varscale end
 Return a vector of scaling factors for the Lagrange multipliers associated with constraints.
 This can be used to improve numerical stability or condition number when solving KKT systems.
 """
-function lagscale end
+function lagscale(model::AbstractNLPModel{T, S}) where {T, S}
+  return ones(T, model.meta.ncon)
+end
 
 """
     conscale(model::AbstractNLPModel)
@@ -1404,4 +1408,6 @@ Return a vector of constraint scaling factors for the model.
 These are typically used to normalize constraints to have similar magnitudes and improve 
 convergence behavior in nonlinear solvers.
 """
-function conscale end
+function conscale(model::AbstractNLPModel{T, S}) where {T, S}
+  return ones(T, model.meta.ncon)
+end
\ No newline at end of file
diff --git a/test/nlp/dense-model.jl b/test/nlp/dense-model.jl
new file mode 100644
index 00000000..8b92dc48
--- /dev/null
+++ b/test/nlp/dense-model.jl
@@ -0,0 +1,250 @@
+"""
+    ManualDenseNLPModel <: AbstractDenseNLPModel
+
+Concrete dense NLP model for demonstration and testing.
+This model stores Jacobian and Hessian as dense matrices.
+
+Example problem:
+    min   x₁² + x₂²
+    s.to  x₁ + x₂ = 1
+          x₁² + x₂² ≤ 2
+
+x₀ = [0.5, 0.5]
+"""
+mutable struct ManualDenseNLPModel{T, S} <: AbstractDenseNLPModel{T, S}
+  meta::NLPModelMeta{T, S}
+  counters::Counters
+end
+
+function ManualDenseNLPModel(::Type{T}) where {T}
+  meta = NLPModelMeta(
+    2,                      # nvar
+    ncon = 2,              # number of constraints
+    nnzj = 4,              # 2x2 dense Jacobian = 4 entries
+    nnzh = 3,              # lower triangle of 2x2 Hessian = 3 entries
+    x0 = T[0.5, 0.5],
+    lcon = T[-Inf, -Inf],
+    ucon = T[0, 0],        # x₁ + x₂ = 0 (shift by -1), x₁² + x₂² ≤ 2
+    name = "Manual Dense NLP Model",
+    lin = [1],             # first constraint is linear
+    lin_nnzj = 2,
+    nln_nnzj = 2,
+  )
+
+  return ManualDenseNLPModel(meta, Counters())
+end
+
+ManualDenseNLPModel() = ManualDenseNLPModel(Float64)
+
+# Objective: f(x) = x₁² + x₂²
+function NLPModels.obj(nlp::ManualDenseNLPModel, x::AbstractVector)
+  @lencheck 2 x
+  increment!(nlp, :neval_obj)
+  return x[1]^2 + x[2]^2
+end
+
+function NLPModels.grad!(nlp::ManualDenseNLPModel, x::AbstractVector, gx::AbstractVector)
+  @lencheck 2 x gx
+  increment!(nlp, :neval_grad)
+  gx[1] = 2 * x[1]
+  gx[2] = 2 * x[2]
+  return gx
+end
+
+# Constraints: c₁(x) = x₁ + x₂ - 1, c₂(x) = x₁² + x₂² - 2
+function NLPModels.cons_lin!(nlp::ManualDenseNLPModel, x::AbstractVector, cx::AbstractVector)
+  @lencheck 2 x
+  @lencheck 1 cx
+  increment!(nlp, :neval_cons_lin)
+  cx[1] = x[1] + x[2] - 1
+  return cx
+end
+
+function NLPModels.cons_nln!(nlp::ManualDenseNLPModel, x::AbstractVector, cx::AbstractVector)
+  @lencheck 2 x
+  @lencheck 1 cx
+  increment!(nlp, :neval_cons_nln)
+  cx[1] = x[1]^2 + x[2]^2 - 2
+  return cx
+end
+
+# Jacobian structure for dense model (all entries)
+function NLPModels.jac_structure!(
+  nlp::ManualDenseNLPModel,
+  rows::AbstractVector{Int},
+  cols::AbstractVector{Int},
+)
+  @lencheck 4 rows cols
+  # For 2 constraints × 2 variables = 4 entries
+  # Row-major order: (1,1), (1,2), (2,1), (2,2)
+  rows .= [1, 1, 2, 2]
+  cols .= [1, 2, 1, 2]
+  return rows, cols
+end
+
+function NLPModels.jac_lin_structure!(
+  nlp::ManualDenseNLPModel,
+  rows::AbstractVector{Int},
+  cols::AbstractVector{Int},
+)
+  @lencheck 2 rows cols
+  # Linear constraint: c₁ = x₁ + x₂ - 1
+  # ∇c₁ = [1, 1]
+  rows .= [1, 1]
+  cols .= [1, 2]
+  return rows, cols
+end
+
+function NLPModels.jac_nln_structure!(
+  nlp::ManualDenseNLPModel,
+  rows::AbstractVector{Int},
+  cols::AbstractVector{Int},
+)
+  @lencheck 2 rows cols
+  # Nonlinear constraint: c₂ = x₁² + x₂² - 2
+  # ∇c₂ = [2x₁, 2x₂]
+  rows .= [1, 1]
+  cols .= [1, 2]
+  return rows, cols
+end
+
+# Jacobian coordinate values
+function NLPModels.jac_coord!(
+  nlp::ManualDenseNLPModel,
+  x::AbstractVector,
+  vals::AbstractVector,
+)
+  @lencheck 2 x
+  @lencheck 4 vals
+  increment!(nlp, :neval_jac)
+  # c₁: [1, 1]
+  vals[1] = 1.0
+  vals[2] = 1.0
+  # c₂: [2x₁, 2x₂]
+  vals[3] = 2 * x[1]
+  vals[4] = 2 * x[2]
+  return vals
+end
+
+function NLPModels.jac_lin_coord!(nlp::ManualDenseNLPModel, vals::AbstractVector)
+  @lencheck 2 vals
+  increment!(nlp, :neval_jac_lin)
+  vals .= 1.0
+  return vals
+end
+
+function NLPModels.jac_nln_coord!(
+  nlp::ManualDenseNLPModel,
+  x::AbstractVector,
+  vals::AbstractVector,
+)
+  @lencheck 2 x vals
+  increment!(nlp, :neval_jac_nln)
+  vals[1] = 2 * x[1]
+  vals[2] = 2 * x[2]
+  return vals
+end
+
+# Jacobian-vector products
+function NLPModels.jprod_lin!(
+  nlp::ManualDenseNLPModel,
+  v::AbstractVector,
+  Jv::AbstractVector,
+)
+  @lencheck 2 v
+  @lencheck 1 Jv
+  increment!(nlp, :neval_jprod_lin)
+  Jv[1] = v[1] + v[2]
+  return Jv
+end
+
+function NLPModels.jprod_nln!(
+  nlp::ManualDenseNLPModel,
+  x::AbstractVector,
+  v::AbstractVector,
+  Jv::AbstractVector,
+)
+  @lencheck 2 x v
+  @lencheck 1 Jv
+  increment!(nlp, :neval_jprod_nln)
+  Jv[1] = 2 * x[1] * v[1] + 2 * x[2] * v[2]
+  return Jv
+end
+
+function NLPModels.jtprod_lin!(
+  nlp::ManualDenseNLPModel,
+  v::AbstractVector,
+  Jtv::AbstractVector,
+)
+  @lencheck 1 v
+  @lencheck 2 Jtv
+  increment!(nlp, :neval_jtprod_lin)
+  Jtv[1] = v[1]
+  Jtv[2] = v[1]
+  return Jtv
+end
+
+function NLPModels.jtprod_nln!(
+  nlp::ManualDenseNLPModel,
+  x::AbstractVector,
+  v::AbstractVector,
+  Jtv::AbstractVector,
+)
+  @lencheck 2 x Jtv
+  @lencheck 1 v
+  increment!(nlp, :neval_jtprod_nln)
+  Jtv[1] = 2 * x[1] * v[1]
+  Jtv[2] = 2 * x[2] * v[1]
+  return Jtv
+end
+
+# Hessian structure (lower triangle)
+function NLPModels.hess_structure!(
+  nlp::ManualDenseNLPModel,
+  rows::AbstractVector{Int},
+  cols::AbstractVector{Int},
+)
+  @lencheck 3 rows cols
+  # Lower triangle of 2×2: (1,1), (2,1), (2,2)
+  rows .= [1, 2, 2]
+  cols .= [1, 1, 2]
+  return rows, cols
+end
+
+# Lagrangian Hessian: ∇²L = obj_weight * ∇²f + y₁ * ∇²c₁ + y₂ * ∇²c₂
+# ∇²f = [2, 0; 0, 2]
+# ∇²c₁ = [0, 0; 0, 0] (linear)
+# ∇²c₂ = [2, 0; 0, 2]
+function NLPModels.hess_coord!(
+  nlp::ManualDenseNLPModel,
+  x::AbstractVector{T},
+  y::AbstractVector{T},
+  vals::AbstractVector{T};
+  obj_weight = one(T),
+) where {T}
+  @lencheck 2 x y
+  @lencheck 3 vals
+  increment!(nlp, :neval_hess)
+  # Lower triangle: (1,1), (2,1), (2,2)
+  vals[1] = 2 * obj_weight + 2 * y[2]  # (1,1)
+  vals[2] = 0                           # (2,1)
+  vals[3] = 2 * obj_weight + 2 * y[2]  # (2,2)
+  return vals
+end
+
+function NLPModels.hprod!(
+  nlp::ManualDenseNLPModel,
+  x::AbstractVector{T},
+  y::AbstractVector{T},
+  v::AbstractVector{T},
+  Hv::AbstractVector{T};
+  obj_weight = one(T),
+) where {T}
+  @lencheck 2 x y v Hv
+  increment!(nlp, :neval_hprod)
+  # H = diag([2*obj_weight + 2*y[2], 2*obj_weight + 2*y[2]])
+  d = 2 * obj_weight + 2 * y[2]
+  Hv[1] = d * v[1]
+  Hv[2] = d * v[2]
+  return Hv
+end
diff --git a/test/nlp/dense.jl b/test/nlp/dense.jl
new file mode 100644
index 00000000..86f81417
--- /dev/null
+++ b/test/nlp/dense.jl
@@ -0,0 +1,53 @@
+using Test
+using NLPModels
+
+@testset "ManualDenseNLPModel dense API" begin
+  model = ManualDenseNLPModel()
+  x = [0.5, 0.5]
+  
+  # Test objective and gradient
+  @test obj(model, x) ≈ 0.5
+  g = similar(x)
+  grad!(model, x, g)
+  @test g ≈ [1.0, 1.0]
+  
+  # Test constraints
+  c = zeros(2)
+  cons!(model, x, c)
+  @test c[1] ≈ 0.0  # x₁ + x₂ - 1 = 0.5 + 0.5 - 1 = 0
+  @test c[2] ≈ -1.5  # x₁² + x₂² - 2 = 0.5 - 2 = -1.5
+  
+  # Test Jacobian structure
+  rows, cols = jac_structure(model)
+  @test length(rows) == 4
+  @test length(cols) == 4
+  @test rows == [1, 1, 2, 2]
+  @test cols == [1, 2, 1, 2]
+  
+  # Test Jacobian values
+  vals = zeros(4)
+  jac_coord!(model, x, vals)
+  @test vals[1] ≈ 1.0   # ∂c₁/∂x₁
+  @test vals[2] ≈ 1.0   # ∂c₁/∂x₂
+  @test vals[3] ≈ 1.0   # ∂c₂/∂x₁ = 2*0.5
+  @test vals[4] ≈ 1.0   # ∂c₂/∂x₂ = 2*0.5
+  
+  # Test Hessian structure
+  rows_h, cols_h = hess_structure(model)
+  @test length(rows_h) == 3
+  @test length(cols_h) == 3
+  @test rows_h == [1, 2, 2]
+  @test cols_h == [1, 1, 2]
+  
+  # Test Hessian values
+  y = [1.0, 1.0]
+  vals_h = zeros(3)
+  hess_coord!(model, x, y, vals_h)
+  @test vals_h[1] ≈ 4.0  # ∇²L₁₁ = 2 (obj) + 2 (y₂)
+  @test vals_h[2] ≈ 0.0  # ∇²L₂₁ = 0
+  @test vals_h[3] ≈ 4.0  # ∇²L₂₂ = 2 (obj) + 2 (y₂)
+  
+  # Test that model is correctly typed as AbstractDenseNLPModel
+  @test model isa AbstractDenseNLPModel
+  @test model isa AbstractNLPModel
+end
diff --git a/test/nlp/dummy-model.jl b/test/nlp/dummy-model.jl
index 4dd2f9b5..4af0f164 100644
--- a/test/nlp/dummy-model.jl
+++ b/test/nlp/dummy-model.jl
@@ -4,16 +4,16 @@ end
 
 @testset "Default methods throw MethodError on DummyModel since they're not defined" begin
   model = DummyModel(NLPModelMeta(1))
-  @test_throws(MethodError, lagscale(model, 1.0))
+  @test lagscale(model) == ones(Float64, model.meta.ncon)
+  @test varscale(model) == ones(Float64, model.meta.nvar)
+  @test conscale(model) == ones(Float64, model.meta.ncon)
   @test_throws(MethodError, obj(model, [0.0]))
-  @test_throws(MethodError, varscale(model, [0.0]))
-  @test_throws(MethodError, conscale(model, [0.0]))
   @test_throws(MethodError, jac_structure(model, [0], [1]))
   @test_throws(MethodError, hess_structure(model, [0], [1]))
   @test_throws(MethodError, grad!(model, [0.0], [1.0]))
   @test_throws(MethodError, cons_lin!(model, [0.0], [1.0]))
   @test_throws(MethodError, cons_nln!(model, [0.0], [1.0]))
-  @test_throws(MethodError, jac_lin_coord!(model, [0.0], [1.0]))
+  @test_throws(MethodError, jac_lin_coord!(model, [1.0]))
   @test_throws(MethodError, jac_nln_coord!(model, [0.0], [1.0]))
   @test_throws(MethodError, jth_con(model, [0.0], 1))
   @test_throws(MethodError, jth_congrad(model, [0.0], 1))
diff --git a/test/runtests.jl b/test/runtests.jl
index 06639cf7..b3d7f0d7 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -2,6 +2,7 @@ using LinearAlgebra, LinearOperators, NLPModels, SparseArrays, Test
 
 include("nlp/simple-model.jl")
 include("nlp/dummy-model.jl")
+include("nlp/dense-model.jl")
 
 include("nlp/api.jl")
 include("nlp/counters.jl")