diff --git a/README.md b/README.md
index e00a21f..7e53a04 100644
--- a/README.md
+++ b/README.md
@@ -98,6 +98,13 @@ const result = await ctx.text2run(`
 models({
   // Set default model when no @model specified
   default: "gpt-4.1-mini",
+
+  // Or set the default model and its request defaults together.
+  // All fields except `model` are added to its outgoing request payload.
+  default: {
+    model: "openai/gpt-5.6-terra",
+    reasoning: { max_tokens: 8000 }
+  },
   
   // Create aliases for models  
   alias: { "gpt": "gpt-4.1-mini" },
@@ -117,8 +124,23 @@ models({
     openai: process.env.OPENAI_KEY,
     anthropic: process.env.ANTHROPIC_KEY,
     ollama: process.env.OLLAMA_URL
+  },
+
+  // Provider request fields used as defaults for individual models.
+  // Keys are real model IDs, so aliases and `default` inherit them.
+  params: {
+    "openai/gpt-5.6-luna": {
+      reasoning: { max_tokens: 8000 }
+    },
+    "claude-sonnet-4-20250514": {
+      thinking: { type: "enabled", budget_tokens: 8000 },
+      max_tokens: 12000
+    }
   }
 })
+
+// `params` are deep-merged into the outgoing request. Fields set later in a
+// chat (for example by `| prop`) override these configured defaults.
 ```
 
 ### Provider-Specific
diff --git a/src/index.js b/src/index.js
index 1e98f84..da79904 100644
--- a/src/index.js
+++ b/src/index.js
@@ -11,16 +11,47 @@ const tune = require("tune-sdk")
 const man = require("tune-sdk/man");
 man.addPackage(__dirname)
 
+// Merge request defaults without replacing nested provider options. Values in
+// `override` win, so inline model processors can override configured defaults.
+function mergeParams(defaults = {}, override = {}) {
+  const result = { ...defaults }
+  for (const [key, value] of Object.entries(override)) {
+    result[key] = value && typeof value === "object" && !Array.isArray(value)
+      ? mergeParams(defaults[key] || {}, value)
+      : value
+  }
+  return result
+}
+
+function withParams(node, defaults) {
+  if (!node || !defaults || node.type !== "llm") return node
+  return {
+    ...node,
+    exec: (payload, ctx) => node.exec(mergeParams(defaults, payload), ctx)
+  }
+}
+
+// `default` may be a model name or `{ model, ...requestDefaults }`.
+function defaultConfig(value) {
+  if (typeof value === "string") return { model: value, params: {} }
+  if (!value || typeof value !== "object") return { model: undefined, params: {} }
+  const { model, ...params } = value
+  return { model, params }
+}
+
 function createModelsMiddleware(options = {}) {
   const { 
     cache = true,           // disk cache by default for text editors
     cacheTtl = 3600000,     // 1 hour default
-    default: defaultModel,
+    default: defaultOption,
     apiKeys = {},
     expose = undefined, 
-    alias = {}
+    alias = {},
+    params = {}
   } = options;
 
+  const { model: defaultModel, params: defaultParams } = defaultConfig(defaultOption)
+
   // Create configured providers
   const providers = [
     ollama({ cache: false, apiKey: apiKeys.ollama }),
@@ -32,14 +63,19 @@ function createModelsMiddleware(options = {}) {
     openrouter({ cache, cacheTtl, apiKey: apiKeys.openrouter }),
   ];
 
+  async function resolveModel(context, modelName, args, requestDefaults = {}) {
+    const node = await tune.resolve(context, modelName, args, providers)
+    return withParams(node, mergeParams(params[modelName], requestDefaults))
+  }
+
   return async function models(name, args) {
     // Handle default model resolution
     if (name === "default" && args.type === "llm" && defaultModel) {
-      return tune.resolve(this, defaultModel, args, providers);
+      return resolveModel(this, defaultModel, args, defaultParams)
     }
 
-
-    // Handle aliases
+    // Resolve aliases before looking up request defaults. This makes aliases
+    // and @default share the configuration of their real model ID.
     const resolvedName = alias[name] || name;
   
     // TODO: what if name is regex?
@@ -47,7 +83,7 @@ function createModelsMiddleware(options = {}) {
       return
     }
 
-    return tune.resolve(this, resolvedName, args, providers);
+    return resolveModel(this, resolvedName, args)
   }
 }
 
@@ -60,4 +96,7 @@ module.exports.gemini = gemini;
 module.exports.mistral = mistral;
 module.exports.groq = groq;
 module.exports.ollama = ollama;
+module.exports.mergeParams = mergeParams;
+module.exports.withParams = withParams;
+module.exports.defaultConfig = defaultConfig;
 
diff --git a/test/index.js b/test/index.js
index 1adfb9d..fafd8e0 100644
--- a/test/index.js
+++ b/test/index.js
@@ -5,7 +5,7 @@ const fs = require('fs');
 const path = require('path');
 const tune = require('tune-sdk');
 const models = require('../src/index');
-const { openai, anthropic, groq, mistral, gemini, openrouter } = require('../src/index');
+const { openai, anthropic, groq, mistral, gemini, openrouter, mergeParams, withParams, defaultConfig } = require('../src/index');
 const llmUtils = require('../src/llm-utils.js')
 
 require('dotenv').config()
@@ -22,6 +22,47 @@ const env = {
 
 const tests = {};
 
+
+tests.params_merge = async function() {
+  const defaults = {
+    reasoning: { max_tokens: 8000, enabled: true },
+    max_tokens: 12000,
+    messages: [{ role: "system", content: "default" }]
+  }
+  const payload = {
+    reasoning: { max_tokens: 2000 },
+    max_tokens: 16000,
+    messages: [{ role: "user", content: "call" }]
+  }
+  const merged = mergeParams(defaults, payload)
+  assert.deepEqual(merged.reasoning, { max_tokens: 2000, enabled: true })
+  assert.equal(merged.max_tokens, 16000)
+  assert.deepEqual(merged.messages, payload.messages)
+  assert.deepEqual(defaults.reasoning, { max_tokens: 8000, enabled: true })
+
+  let received
+  const node = withParams({
+    type: "llm",
+    exec: async (request) => {
+      received = request
+      return "request created"
+    }
+  }, defaults)
+  assert.equal(await node.exec(payload), "request created")
+  assert.deepEqual(received, merged)
+
+  assert.deepEqual(defaultConfig("gpt-4.1-mini"), {
+    model: "gpt-4.1-mini", params: {}
+  })
+  assert.deepEqual(defaultConfig({
+    model: "openai/gpt-5.6-terra",
+    reasoning: { max_tokens: 8000 }
+  }), {
+    model: "openai/gpt-5.6-terra",
+    params: { reasoning: { max_tokens: 8000 } }
+  })
+}
+
 tests.api_keys = async function(){
   assert.ok(process.env.OPENAI_KEY, "OPENAI_KEY has to be set for testing") 
   assert.ok(process.env.ANTHROPIC_KEY, "ANTHROPIC_KEY has to be set for testing")