diff --git a/local/ai-provider/lua/ai-provider/core.lua b/local/ai-provider/lua/ai-provider/core.lua
index 67d8b41..47eee0e 100644
--- a/local/ai-provider/lua/ai-provider/core.lua
+++ b/local/ai-provider/lua/ai-provider/core.lua
@@ -1,4 +1,5 @@
 local M = {}
+local log = require 'ai-provider.log'
 
 local providers = {
   ollama = require 'ai-provider.providers.ollama',
@@ -215,6 +216,7 @@ function M.chat(first, second)
   local provider, request = normalize_chat_args(first, second)
   local implementation = M.get_provider(provider)
   if not implementation or not implementation.chat then
+    log.error('chat requested unavailable provider: ' .. tostring(provider))
     if request.callback then
       request.callback(nil, nil)
     end
@@ -224,6 +226,7 @@ function M.chat(first, second)
   request.model = request.model or M.get_selected_model(provider)
   request.provider_config = get_provider_config(provider)
   if not request.model then
+    log.error('chat requested without selected model for provider: ' .. tostring(provider))
     if request.callback then
       request.callback(nil, nil)
     end
@@ -231,6 +234,17 @@ function M.chat(first, second)
     return nil
   end
 
+  log.info(
+    string.format(
+      'chat start provider=%s model=%s prompt_chars=%d max_tokens=%s context_size=%s stream=%s',
+      provider,
+      request.model,
+      type(request.prompt) == 'string' and #request.prompt or 0,
+      tostring(request.max_tokens),
+      tostring(request.context_size),
+      tostring(request.stream ~= false)
+    )
+  )
   return implementation.chat(request)
 end
 
@@ -334,6 +348,45 @@ function M.select_model(provider)
   end, { force = true })
 end
 
+function M.select_helper(opts, callback)
+  if type(opts) == 'function' then
+    callback = opts
+    opts = {}
+  end
+  opts = opts or {}
+
+  collect_models(function(models)
+    if #models == 0 then
+      vim.notify('No AI provider models are available.', vim.log.levels.WARN)
+      if callback then
+        callback(nil)
+      end
+      return
+    end
+
+    local current = opts.current
+    vim.ui.select(models, {
+      prompt = opts.prompt or 'Select AI provider model:',
+      format_item = function(item)
+        local selected = current and item.provider == current.provider and item.model == current.model
+        local marker = selected and '✓ ' or '  '
+        return marker .. item.label
+      end,
+    }, function(choice)
+      if not choice then
+        if callback then
+          callback(nil)
+        end
+        return
+      end
+
+      if callback then
+        callback({ provider = choice.provider, model = choice.model, label = choice.label })
+      end
+    end)
+  end)
+end
+
 local global_actions = { 'default', 'model', 'models' }
 local provider_actions = { 'auth', 'check', 'model', 'models' }
 
diff --git a/local/ai-provider/lua/ai-provider/init.lua b/local/ai-provider/lua/ai-provider/init.lua
index 4a6b021..1b91665 100644
--- a/local/ai-provider/lua/ai-provider/init.lua
+++ b/local/ai-provider/lua/ai-provider/init.lua
@@ -9,11 +9,23 @@
 ---@field timeout? integer Optional provider timeout in milliseconds.
 ---@field context_size? integer Optional provider-wide default context size.
 ---@field keep_alive? string|integer Optional provider-wide keep-alive/unload timeout. Ollama accepts values like `4h`, `10m`, or `0`.
+---@field load_timeout? integer Optional model preload timeout in milliseconds. Used by Ollama before the normal chat timeout starts.
+---@field think? boolean Optional Ollama thinking mode override for reasoning models.
 ---@field models? table<string, string|AiProviderModelConfig> Optional logical model profiles. Keys are selectable model names.
 
+---@class AiProviderHelperConfig
+---@field provider string Provider name used by the helper.
+---@field model string Model/profile name returned by the helper. This is a reference to the configured model name, not copied options.
+---@field label? string Display label, usually `provider/model`.
+
+---@class AiProviderSelectHelperOptions
+---@field prompt? string Picker prompt.
+---@field current? AiProviderHelperConfig Current selection used for picker highlighting.
+
 ---@class AiProviderModelConfig
 ---@field model string Underlying provider model name.
 ---@field context_size? integer Optional model-specific context size override.
+---@field think? boolean Optional Ollama thinking mode override for this logical model profile.
 
 ---@class AiProviderConfig
 ---@field default_provider string Default provider used when no persisted default provider exists.
@@ -23,6 +35,23 @@
 ---@field requested_model string Model requested by the caller.
 ---@field used_model string Model reported by the provider response.
 ---@field elapsed_ms number Request duration in milliseconds.
+---@field done_reason? string Provider stop reason, if reported.
+---@field error? string Provider/runtime error when `message` is nil.
+---@field total_duration? integer Provider total duration in nanoseconds, when reported.
+---@field load_duration? integer Initial model load duration in nanoseconds, when reported.
+---@field prompt_eval_count? integer Prompt tokens evaluated, when reported.
+---@field prompt_eval_duration? integer Prompt evaluation duration in nanoseconds, when reported.
+---@field eval_count? integer Generated tokens evaluated, when reported.
+---@field eval_duration? integer Generation duration in nanoseconds, when reported.
+
+---@class AiProviderStatus
+---@field provider string Provider name.
+---@field phase string Standard phase, for example `loading`, `loaded`, `thinking`, `generating`, `done`, or `error`.
+---@field message string Human-readable status message.
+---@field model? string Model/profile name.
+---@field used_model? string Raw provider model name.
+---@field tokens? integer Token count when the provider reports one.
+---@field elapsed_ms? number Elapsed duration in milliseconds.
 
 ---@class AiProviderChatRequest
 ---@field provider? string Provider name. Defaults to `get_default_provider()`.
@@ -32,7 +61,11 @@
 ---@field max_tokens? integer Maximum generated tokens/provider equivalent.
 ---@field context_size? integer Per-request context size override.
 ---@field keep_alive? string|integer Per-request keep-alive/unload timeout override.
+---@field load_timeout? integer Per-request model preload timeout in milliseconds.
+---@field preload? boolean Whether to preload the model before chat. Defaults to false; callers such as ai-commit can enable it to exclude load time from chat timeout.
+---@field think? boolean Per-request Ollama thinking mode override.
 ---@field on_chunk? fun(chunk: string, raw: table) Called for each streamed text chunk.
+---@field on_status? fun(status: AiProviderStatus) Called with standardized provider progress updates.
 ---@field callback? fun(message: string|nil, meta: AiProviderChatMeta|nil) Called once the request finishes.
 ---@field is_cancelled? fun(): boolean Optional cancellation predicate.
 ---@field register_http_job? fun(job: table) Receives the provider job/process for external cancellation.
@@ -166,6 +199,14 @@ function M.select_model(provider)
   return core.select_model(provider)
 end
 
+---Open a model picker for feature-specific callers. This does not persist state;
+---callers should save the returned provider/model reference themselves.
+---@param opts? AiProviderSelectHelperOptions|fun(selection: AiProviderHelperConfig|nil)
+---@param callback? fun(selection: AiProviderHelperConfig|nil) Called with the selected provider/model reference.
+function M.select_helper(opts, callback)
+  return core.select_helper(opts, callback)
+end
+
 ---@param arglead string Current command-line argument prefix.
 ---@param cmdline string Full command-line.
 ---@return string[] completions Matching command completions.
diff --git a/local/ai-provider/lua/ai-provider/providers/ollama.lua b/local/ai-provider/lua/ai-provider/providers/ollama.lua
index cea9c83..274a4cc 100644
--- a/local/ai-provider/lua/ai-provider/providers/ollama.lua
+++ b/local/ai-provider/lua/ai-provider/providers/ollama.lua
@@ -1,8 +1,10 @@
 local M = {}
 local curl = require 'ai-provider.curl'
+local log = require 'ai-provider.log'
 
 local DEFAULT_ENDPOINT = 'http://127.0.0.1:11434'
 local HEALTH_CACHE_TTL = 30
+local DEFAULT_LOAD_TIMEOUT = 120000
 
 local state = {
   health = nil,
@@ -45,6 +47,17 @@ local function model_config(provider_config, model)
   return { model = model }
 end
 
+local function elapsed_ms_since(started_at)
+  return (vim.uv.hrtime() - started_at) / 1e6
+end
+
+local function tokens_per_second(count, duration_ns)
+  if type(count) ~= 'number' or type(duration_ns) ~= 'number' or duration_ns <= 0 then
+    return nil
+  end
+  return count / (duration_ns / 1e9)
+end
+
 function M.check(callback, opts)
   opts = opts or {}
   local now = os.time()
@@ -59,6 +72,7 @@ function M.check(callback, opts)
     callback = function(response)
       state.health = response.status == 200
       state.health_checked_at = os.time()
+      log.debug('ollama check status=' .. tostring(response.status) .. ' working=' .. tostring(state.health))
       callback(state.health)
     end,
   }
@@ -109,7 +123,53 @@ function M.chat(request)
   local raw_model = selected_config.model or selected_model
   local context_size = request.context_size or selected_config.context_size or provider_config.context_size
   local keep_alive = request.keep_alive or provider_config.keep_alive
+  local think = request.think
+  if think == nil then
+    think = selected_config.think
+  end
+  if think == nil then
+    think = provider_config.think
+  end
   local final_model = raw_model
+  local done_reason = nil
+  local provider_error = nil
+  local metrics = {}
+  local thinking_chars = 0
+  local last_status_key = nil
+
+  local function emit_status(status)
+    if not request.on_status then
+      return
+    end
+
+    status.provider = 'ollama'
+    status.model = status.model or selected_model
+    status.used_model = status.used_model or final_model
+    status.elapsed_ms = status.elapsed_ms or elapsed_ms_since(started_at)
+    local key = table.concat({ status.phase or '', status.message or '', tostring(status.tokens), tostring(status.used_model) }, '|')
+    if key == last_status_key then
+      return
+    end
+    last_status_key = key
+    vim.schedule(function()
+      request.on_status(status)
+    end)
+  end
+
+  log.info(
+    string.format(
+      'ollama request selected_model=%s raw_model=%s prompt_chars=%d context_size=%s max_tokens=%s keep_alive=%s timeout=%s stream=%s think=%s',
+      tostring(selected_model),
+      tostring(raw_model),
+      type(request.prompt) == 'string' and #request.prompt or 0,
+      tostring(context_size),
+      tostring(request.max_tokens),
+      tostring(keep_alive),
+      tostring(request.timeout or 30000),
+      tostring(request.stream ~= false),
+      tostring(think)
+    )
+  )
 
   local body = {
     model = raw_model,
@@ -121,46 +181,285 @@ function M.chat(request)
     },
     keep_alive = keep_alive,
   }
+  if think ~= nil then
+    body.think = think
+  end
 
-  local job = curl.stream_json_lines {
-    url = endpoint() .. '/api/chat',
-    body = body,
-    timeout = request.timeout or 30000,
-    is_cancelled = request.is_cancelled,
-    on_json_line = function(data)
-      if is_cancelled(request) then
-        return
-      end
+  local function run_chat()
+    return curl.stream_json_lines {
+      url = endpoint() .. '/api/chat',
+      body = body,
+      timeout = request.timeout or 30000,
+      is_cancelled = request.is_cancelled,
+      on_json_line = function(data)
+        if is_cancelled(request) then
+          return
+        end
 
-      final_model = data.model or final_model
-      local chunk = data.message and data.message.content or ''
-      if chunk ~= '' then
-        table.insert(chunks, chunk)
-        if request.on_chunk then
-          vim.schedule(function()
-            request.on_chunk(chunk, data)
-          end)
+        final_model = data.model or final_model
+        done_reason = data.done_reason or done_reason
+        if type(data.error) == 'string' and data.error ~= '' then
+          provider_error = data.error
+        end
+        metrics.total_duration = data.total_duration or metrics.total_duration
+        metrics.load_duration = data.load_duration or metrics.load_duration
+        metrics.prompt_eval_count = data.prompt_eval_count or metrics.prompt_eval_count
+        metrics.prompt_eval_duration = data.prompt_eval_duration or metrics.prompt_eval_duration
+        metrics.eval_count = data.eval_count or metrics.eval_count
+        metrics.eval_duration = data.eval_duration or metrics.eval_duration
+        local thinking = data.message and data.message.thinking or ''
+        if thinking ~= '' then
+          thinking_chars = thinking_chars + #thinking
+          emit_status {
+            phase = 'thinking',
+            message = 'Thinking',
+            tokens = data.eval_count,
+          }
+        end
+        local chunk = data.message and data.message.content or ''
+        if chunk ~= '' then
+          table.insert(chunks, chunk)
+          emit_status {
+            phase = thinking_chars > 0 and 'generating' or 'generating',
+            message = 'Generating response',
+            tokens = data.eval_count,
+          }
+          if request.on_chunk then
+            vim.schedule(function()
+              request.on_chunk(chunk, data)
+            end)
+          end
+        end
+      end,
+      callback = function(code, error_message)
+        if is_cancelled(request) then
+          return
+        end
+
+        if code ~= 0 then
+          log.error(
+            'ollama request process failed code='
+              .. tostring(code)
+              .. ' model='
+              .. tostring(raw_model)
+              .. ' error='
+              .. tostring(error_message)
+          )
+          if request.callback then
+            request.callback(nil, {
+              requested_model = selected_model,
+              used_model = final_model,
+              elapsed_ms = (vim.uv.hrtime() - started_at) / 1e6,
+              error = error_message or 'ollama request failed',
+            })
+          end
+          emit_status {
+            phase = 'error',
+            message = error_message or 'Ollama request failed',
+          }
+          return
+        end
+
+        local elapsed_ms = (vim.uv.hrtime() - started_at) / 1e6
+        local meta = vim.tbl_extend('force', {
+          requested_model = selected_model,
+          used_model = final_model,
+          elapsed_ms = elapsed_ms,
+          done_reason = done_reason,
+        }, metrics)
+        local message = table.concat(chunks, '')
+        local prompt_tokens_per_second = tokens_per_second(metrics.prompt_eval_count, metrics.prompt_eval_duration)
+        local eval_tokens_per_second = tokens_per_second(metrics.eval_count, metrics.eval_duration)
+        log.info(
+          string.format(
+            'ollama response requested_model=%s used_model=%s done_reason=%s elapsed_ms=%.0f output_chars=%d context_size=%s max_tokens=%s load_ms=%s prompt_eval_count=%s prompt_eval_ms=%s prompt_tokens_per_second=%s eval_count=%s eval_ms=%s tokens_per_second=%s total_ms=%s',
+            tostring(selected_model),
+            tostring(final_model),
+            tostring(done_reason),
+            elapsed_ms,
+            #message,
+            tostring(context_size),
+            tostring(request.max_tokens),
+            metrics.load_duration and string.format('%.0f', metrics.load_duration / 1e6) or 'nil',
+            tostring(metrics.prompt_eval_count),
+            metrics.prompt_eval_duration and string.format('%.0f', metrics.prompt_eval_duration / 1e6) or 'nil',
+            prompt_tokens_per_second and string.format('%.2f', prompt_tokens_per_second) or 'nil',
+            tostring(metrics.eval_count),
+            metrics.eval_duration and string.format('%.0f', metrics.eval_duration / 1e6) or 'nil',
+            eval_tokens_per_second and string.format('%.2f', eval_tokens_per_second) or 'nil',
+            metrics.total_duration and string.format('%.0f', metrics.total_duration / 1e6) or 'nil'
+          )
+        )
+        if provider_error then
+          meta.error = provider_error
+          log.error('ollama provider error requested_model=' .. tostring(selected_model) .. ' error=' .. provider_error)
+          if request.callback then
+            request.callback(nil, meta)
+          end
+          emit_status {
+            phase = 'error',
+            message = provider_error,
+            tokens = metrics.eval_count,
+          }
+          return
+        end
+        if done_reason == 'length' then
+          meta.error = 'ollama stopped because the context or generation length limit was reached'
+          local prompt_tokens_per_second = tokens_per_second(metrics.prompt_eval_count, metrics.prompt_eval_duration)
+          local eval_tokens_per_second = tokens_per_second(metrics.eval_count, metrics.eval_duration)
+          log.error(
+            string.format(
+              'ollama length stop requested_model=%s used_model=%s prompt_chars=%d output_chars=%d context_size=%s max_tokens=%s load_ms=%s prompt_eval_count=%s prompt_eval_ms=%s prompt_tokens_per_second=%s eval_count=%s eval_ms=%s tokens_per_second=%s',
+              tostring(selected_model),
+              tostring(final_model),
+              type(request.prompt) == 'string' and #request.prompt or 0,
+              #message,
+              tostring(context_size),
+              tostring(request.max_tokens),
+              metrics.load_duration and string.format('%.0f', metrics.load_duration / 1e6) or 'nil',
+              tostring(metrics.prompt_eval_count),
+              metrics.prompt_eval_duration and string.format('%.0f', metrics.prompt_eval_duration / 1e6) or 'nil',
+              prompt_tokens_per_second and string.format('%.2f', prompt_tokens_per_second) or 'nil',
+              tostring(metrics.eval_count),
+              metrics.eval_duration and string.format('%.0f', metrics.eval_duration / 1e6) or 'nil',
+              eval_tokens_per_second and string.format('%.2f', eval_tokens_per_second) or 'nil'
+            )
+          )
+          if request.callback then
+            request.callback(nil, meta)
+          end
+          emit_status {
+            phase = 'error',
+            message = meta.error,
+            tokens = metrics.eval_count,
+          }
+          return
         end
-      end
-    end,
-    callback = function(code)
-      if is_cancelled(request) then
-        return
-      end
 
-      if code ~= 0 then
+        if message == '' then
+          meta.error = 'ollama returned no content'
+          log.error('ollama returned no content requested_model=' .. tostring(selected_model) .. ' done_reason=' .. tostring(done_reason))
+          if request.callback then
+            request.callback(nil, meta)
+          end
+          emit_status {
+            phase = 'error',
+            message = meta.error,
+            tokens = metrics.eval_count,
+          }
+          return
+        end
+
+        emit_status {
+          phase = 'done',
+          message = 'Response complete',
+          tokens = metrics.eval_count,
+        }
         if request.callback then
-          request.callback(nil, nil)
+          request.callback(message, meta)
         end
-        return
-      end
+      end,
+    }
+  end
 
-      local elapsed_ms = (vim.uv.hrtime() - started_at) / 1e6
-      if request.callback then
-        request.callback(table.concat(chunks, ''), { requested_model = selected_model, used_model = final_model, elapsed_ms = elapsed_ms })
-      end
-    end,
-  }
+  local job = nil
+  local load_timeout = request.load_timeout or provider_config.load_timeout or DEFAULT_LOAD_TIMEOUT
+  if request.preload == true or provider_config.preload == true then
+    emit_status {
+      phase = 'loading',
+      message = 'Loading model',
+    }
+    log.info(
+      string.format(
+        'ollama preload start selected_model=%s raw_model=%s context_size=%s keep_alive=%s load_timeout=%s',
+        tostring(selected_model),
+        tostring(raw_model),
+        tostring(context_size),
+        tostring(keep_alive),
+        tostring(load_timeout)
+      )
+    )
+    local preload_body = {
+      model = raw_model,
+      messages = { { role = 'user', content = 'ok' } },
+      stream = false,
+      keep_alive = keep_alive,
+      options = {
+        num_ctx = context_size,
+        num_predict = 16,
+      },
+    }
+    if think ~= nil then
+      preload_body.think = think
+    end
+
+    job = curl.json {
+      method = 'POST',
+      url = endpoint() .. '/api/chat',
+      timeout = load_timeout,
+      body = preload_body,
+      callback = function(response)
+        if is_cancelled(request) then
+          return
+        end
+
+        if response.status ~= 200 then
+          local error_message = response.error or response.body or 'ollama preload failed'
+          log.error(
+            'ollama preload failed status='
+              .. tostring(response.status)
+              .. ' model='
+              .. tostring(raw_model)
+              .. ' error='
+              .. tostring(error_message)
+          )
+          if request.callback then
+            request.callback(nil, {
+              requested_model = selected_model,
+              used_model = raw_model,
+              elapsed_ms = (vim.uv.hrtime() - started_at) / 1e6,
+              error = error_message,
+            })
+          end
+          emit_status {
+            phase = 'error',
+            message = error_message,
+          }
+          return
+        end
+
+        local load_duration = type(response.json) == 'table' and response.json.load_duration or nil
+        emit_status {
+          phase = 'loaded',
+          message = 'Model loaded',
+          elapsed_ms = load_duration and (load_duration / 1e6) or elapsed_ms_since(started_at),
+        }
+        log.info(
+          string.format(
+            'ollama preload complete selected_model=%s raw_model=%s status=%s load_ms=%s',
+            tostring(selected_model),
+            tostring(raw_model),
+            tostring(response.status),
+            load_duration and string.format('%.0f', load_duration / 1e6) or 'nil'
+          )
+        )
+        emit_status {
+          phase = 'generating',
+          message = 'Generating response',
+        }
+        job = run_chat()
+        if request.register_http_job then
+          request.register_http_job(job)
+        end
+      end,
+    }
+  else
+    emit_status {
+      phase = 'generating',
+      message = 'Generating response',
+    }
+    job = run_chat()
+  end
 
   if request.register_http_job then
     request.register_http_job(job)
diff --git a/local/ai-provider/tests/ai_provider/core_spec.lua b/local/ai-provider/tests/ai_provider/core_spec.lua
index 7cbe480..20d4588 100644
--- a/local/ai-provider/tests/ai_provider/core_spec.lua
+++ b/local/ai-provider/tests/ai_provider/core_spec.lua
@@ -26,4 +26,35 @@ describe('ai-provider core', function()
     assert.is_false(ok)
     assert.matches('default_provider must be configured', err)
   end)
+
+  it('select helper returns a provider/model reference without saving it globally', function()
+    ai_provider.setup {
+      default_provider = 'ollama',
+      providers = {
+        ollama = { default_model = 'gemma4:e2b' },
+      },
+    }
+
+    local core = require 'ai-provider.core'
+    local original_list_models = core.list_models
+    local original_select = vim.ui.select
+    local selected = nil
+
+    rawset(core, 'list_models', function(provider, callback)
+      callback(provider == 'ollama' and { 'gemma4:e2b 64k' } or {})
+    end)
+    rawset(vim.ui, 'select', function(items, _, on_choice)
+      on_choice(items[1])
+    end)
+
+    ai_provider.select_helper(function(choice)
+      selected = choice
+    end)
+
+    rawset(core, 'list_models', original_list_models)
+    rawset(vim.ui, 'select', original_select)
+
+    assert.are.same({ provider = 'ollama', model = 'gemma4:e2b 64k', label = 'ollama/gemma4:e2b 64k' }, selected)
+    assert.are.same('gemma4:e2b', ai_provider.get_selected_model 'ollama')
+  end)
 end)
diff --git a/local/ai-provider/tests/ai_provider/ollama_options_spec.lua b/local/ai-provider/tests/ai_provider/ollama_options_spec.lua
index eca1248..e4bbd81 100644
--- a/local/ai-provider/tests/ai_provider/ollama_options_spec.lua
+++ b/local/ai-provider/tests/ai_provider/ollama_options_spec.lua
@@ -30,6 +30,7 @@ describe('ollama provider options', function()
     ollama.chat {
       model = 'gemma4:e2b 64k',
       prompt = 'Reply with exactly: ok',
+      preload = false,
       max_tokens = 4,
       provider_config = {
         context_size = 1024 * 8,
@@ -74,6 +75,7 @@ describe('ollama provider options', function()
     ollama.chat {
       model = 'gemma4:e2b 32k',
       prompt = 'Reply with exactly: ok',
+      preload = false,
       context_size = 1024 * 16,
       keep_alive = '10m',
       provider_config = {
@@ -95,4 +97,107 @@ describe('ollama provider options', function()
     assert.are.same(1024 * 16, captured.options.num_ctx)
     assert.are.same('10m', captured.keep_alive)
   end)
+
+  it('sends configured thinking mode for reasoning model profiles', function()
+    ---@type table|nil
+    local captured = nil
+    package.loaded['ai-provider.curl'] = {
+      stream_json_lines = function(request)
+        captured = request.body
+        request.on_json_line { model = 'qwen3.5:4b', message = { content = 'ok' }, done_reason = 'stop' }
+        request.callback(0)
+        return { shutdown = function() end }
+      end,
+    }
+
+    local ollama = require 'ai-provider.providers.ollama'
+
+    ollama.chat {
+      model = 'qwen3.5:4b 256k',
+      prompt = 'Reply with exactly: ok',
+      preload = false,
+      provider_config = {
+        models = {
+          ['qwen3.5:4b 256k'] = {
+            model = 'qwen3.5:4b',
+            context_size = 1024 * 256,
+            think = false,
+          },
+        },
+      },
+      callback = function() end,
+    }
+
+    assert.is_table(captured)
+    ---@cast captured table
+    assert.are.same('qwen3.5:4b', captured.model)
+    assert.are.same(1024 * 256, captured.options.num_ctx)
+    assert.are.same(false, captured.think)
+  end)
+
+  it('returns an error instead of partial output when ollama stops for length', function()
+    package.loaded['ai-provider.curl'] = {
+      stream_json_lines = function(request)
+        request.on_json_line { model = 'gemma4:e2b', message = { content = 'Ref' }, done_reason = 'length' }
+        request.callback(0)
+        return { shutdown = function() end }
+      end,
+    }
+
+    local ollama = require 'ai-provider.providers.ollama'
+    local message = 'unset'
+    local meta = nil
+
+    ollama.chat {
+      model = 'gemma4:e2b',
+      prompt = 'large prompt',
+      preload = false,
+      callback = function(result, result_meta)
+        message = result
+        meta = result_meta
+      end,
+    }
+
+    assert.is_nil(message)
+    assert.is_table(meta)
+    ---@cast meta table
+    assert.are.same('length', meta.done_reason)
+    assert.matches('length limit', meta.error)
+  end)
+
+  it('emits standardized thinking and generating status events', function()
+    package.loaded['ai-provider.curl'] = {
+      stream_json_lines = function(request)
+        request.on_json_line { model = 'gemma4:e2b', message = { thinking = 'thinking...' }, eval_count = 7 }
+        request.on_json_line { model = 'gemma4:e2b', message = { content = 'ok' }, eval_count = 9 }
+        request.on_json_line { model = 'gemma4:e2b', done_reason = 'stop', eval_count = 9 }
+        request.callback(0)
+        return { shutdown = function() end }
+      end,
+    }
+
+    local ollama = require 'ai-provider.providers.ollama'
+    local statuses = {}
+
+    ollama.chat {
+      model = 'gemma4:e2b',
+      prompt = 'Reply with exactly: ok',
+      preload = false,
+      on_status = function(status)
+        table.insert(statuses, status)
+      end,
+      callback = function() end,
+    }
+
+    vim.wait(1000, function()
+      return #statuses >= 4
+    end, 10)
+
+    assert.are.same('generating', statuses[1].phase)
+    assert.are.same('thinking', statuses[2].phase)
+    assert.are.same(7, statuses[2].tokens)
+    assert.are.same('generating', statuses[3].phase)
+    assert.are.same(9, statuses[3].tokens)
+    assert.are.same('done', statuses[4].phase)
+  end)
 end)
diff --git a/local/ai-provider/tests/ai_provider/ollama_spec.lua b/local/ai-provider/tests/ai_provider/ollama_spec.lua
index a23cd8d..e55fb86 100644
--- a/local/ai-provider/tests/ai_provider/ollama_spec.lua
+++ b/local/ai-provider/tests/ai_provider/ollama_spec.lua
@@ -28,6 +28,27 @@ local function wait_for(done, timeout)
   assert.is_true(ok)
 end
 
+local function run_command(command)
+  local result = vim.system(command, { text = true }):wait()
+  assert.are.same(0, result.code)
+  return result.stdout or ''
+end
+
+local function unload_ollama_model(model)
+  run_command {
+    'curl',
+    '--silent',
+    '--show-error',
+    '--request',
+    'POST',
+    '--header',
+    'Content-Type: application/json',
+    '--data',
+    string.format('{"model":"%s","keep_alive":0}', model),
+    'http://127.0.0.1:11434/api/generate',
+  }
+end
+
 describe('ollama provider integration', function()
   before_each(function()
     ai_provider.setup(config)
@@ -78,7 +99,7 @@ describe('ollama provider integration', function()
     ai_provider.chat('ollama', {
       model = 'gemma4:e2b 32k',
       prompt = 'Reply with exactly: ok',
-      max_tokens = 4,
+      max_tokens = 256,
       timeout = 10000,
       callback = function(result, result_meta)
         message = result
@@ -105,7 +126,7 @@ describe('ollama provider integration', function()
     ai_provider.chat('ollama', {
       model = 'gemma4:e2b',
       prompt = 'Reply with exactly: ok',
-      max_tokens = 4,
+      max_tokens = 16,
       timeout = 10000,
       on_chunk = function(chunk)
         table.insert(chunks, chunk)
@@ -123,4 +144,41 @@ describe('ollama provider integration', function()
     assert.is_true(#chunks > 0)
     assert.are.same('ok', table.concat(chunks, ''))
   end)
+
+  it('returns an error when the prompt exceeds a small loaded context window', function()
+    unload_ollama_model 'gemma4:e2b'
+
+    local done = false
+    ---@type string|nil
+    local message = 'unset'
+    ---@type table|nil
+    local meta = nil
+    local prompt = table.concat(vim.fn['repeat']({ 'context-overflow-token' }, 5000), ' ')
+
+    ai_provider.chat('ollama', {
+      model = 'gemma4:e2b',
+      prompt = prompt,
+      context_size = 2048,
+      max_tokens = 32,
+      timeout = 60000,
+      callback = function(result, result_meta)
+        message = result
+        meta = result_meta
+        done = true
+      end,
+    })
+
+    wait_for(function()
+      return done
+    end, 70000)
+
+    local ps = run_command { 'ollama', 'ps' }
+    assert.matches('gemma4:e2b', ps)
+    assert.matches('%s2048%s', ps)
+    assert.is_nil(message)
+    assert.is_table(meta)
+    ---@cast meta table
+    assert.are.same('length', meta.done_reason)
+    assert.matches('length limit', meta.error)
+  end)
 end)
diff --git a/lua/plugins/ai-commit.lua b/lua/plugins/ai-commit.lua
index 4eae0d7..a6b931c 100644
--- a/lua/plugins/ai-commit.lua
+++ b/lua/plugins/ai-commit.lua
@@ -8,6 +8,9 @@ local CONFIG = {
   -- Use :AIProvider to configure the local provider; this is only the remote fallback model.
   model = nil, -- nil = auto (use Copilot's default)
   model_name = nil, -- Friendly display name for selected model
+  local_provider = 'ollama',
+  local_model = 'gemma4:e2b 64k',
+  local_model_name = 'gemma4:e2b 64k',
 
   openrouter = {
     endpoint = 'https://openrouter.ai/api/v1',
@@ -116,11 +119,20 @@ local function load_preferences()
     if ok and type(prefs) == 'table' then
       CONFIG.provider = prefs.provider or 'copilot'
       CONFIG.model = prefs.model
+      CONFIG.local_provider = prefs.local_provider or CONFIG.local_provider
+      CONFIG.local_model = prefs.local_model or CONFIG.local_model
       if type(prefs.model_name) == 'string' and prefs.model_name ~= '' then
         CONFIG.model_name = prefs.model_name
       else
         CONFIG.model_name = nil
       end
+      if type(prefs.local_model_name) == 'string' and prefs.local_model_name ~= '' then
+        CONFIG.local_model_name = prefs.local_model_name
+      elseif type(CONFIG.local_model) == 'string' and CONFIG.local_model ~= '' then
+        CONFIG.local_model_name = CONFIG.local_model
+      else
+        CONFIG.local_model_name = nil
+      end
     end
   end
 end
@@ -136,6 +148,9 @@ local function save_preferences()
     provider = CONFIG.provider,
     model = CONFIG.model,
     model_name = CONFIG.model_name,
+    local_provider = CONFIG.local_provider,
+    local_model = CONFIG.local_model,
+    local_model_name = CONFIG.local_model_name,
   }
   local file = io.open(prefs_file, 'w')
   if file then
@@ -225,11 +240,12 @@ end
 local function complete_ollama(full_prompt, callback, status_callback, request_context)
   local log = setup_logger()
   local ai_provider = require 'ai-provider'
-  local model = ai_provider.get_selected_model 'ollama'
+  local provider = CONFIG.local_provider or 'ollama'
+  local model = CONFIG.local_model or ai_provider.get_selected_model(provider)
 
   if not model then
-    log.error 'Ollama is reachable but no model is selected'
-    vim.notify('No Ollama model selected. Run :AIProvider ollama model first.', vim.log.levels.ERROR)
+    log.error(provider .. ' is reachable but no model is selected')
+    vim.notify('No ' .. provider .. ' model selected. Run AI commit model selection first.', vim.log.levels.ERROR)
     callback(nil, nil)
     return
   end
@@ -238,21 +254,60 @@ local function complete_ollama(full_prompt, callback, status_callback, request_c
     status_callback('Waiting for response from ' .. model)
   end
 
-  ai_provider.chat('ollama', {
+  local function report_provider_status(status)
+    if not status_callback or type(status) ~= 'table' then
+      return
+    end
+
+    local status_model = status.model or model
+    if status.phase == 'loading' then
+      status_callback('Loading model ' .. status_model)
+    elseif status.phase == 'loaded' then
+      status_callback('Loaded model ' .. status_model)
+    elseif status.phase == 'thinking' then
+      local token_suffix = status.tokens and (' (' .. status.tokens .. ' tokens)') or ''
+      status_callback('Thinking with ' .. status_model .. token_suffix)
+    elseif status.phase == 'generating' then
+      local token_suffix = status.tokens and (' (' .. status.tokens .. ' tokens)') or ''
+      status_callback('Generating response with ' .. status_model .. token_suffix)
+    elseif status.phase == 'error' then
+      status_callback('Provider error from ' .. status_model)
+    end
+  end
+
+  ai_provider.chat(provider, {
     model = model,
     prompt = full_prompt,
     stream = true,
+    preload = true,
     max_tokens = CONFIG.max_tokens,
     is_cancelled = request_context and request_context.is_cancelled,
     register_http_job = request_context and request_context.register_http_job,
+    on_status = report_provider_status,
     callback = function(message, meta)
       if request_context and request_context.is_cancelled and request_context.is_cancelled() then
         return
       end
 
       if not message then
-        log.error 'Ollama chat request failed'
-        vim.notify('Ollama request failed. See ai-commit logs.', vim.log.levels.ERROR)
+        local error_message = meta and meta.error or 'unknown error'
+        local details = ''
+        if meta then
+          details = string.format(
+            ' (requested_model=%s used_model=%s done_reason=%s elapsed_ms=%s load_ms=%s prompt_eval_count=%s prompt_eval_ms=%s eval_count=%s eval_ms=%s)',
+            tostring(meta.requested_model),
+            tostring(meta.used_model),
+            tostring(meta.done_reason),
+            tostring(meta.elapsed_ms),
+            meta.load_duration and string.format('%.0f', meta.load_duration / 1e6) or 'nil',
+            tostring(meta.prompt_eval_count),
+            meta.prompt_eval_duration and string.format('%.0f', meta.prompt_eval_duration / 1e6) or 'nil',
+            tostring(meta.eval_count),
+            meta.eval_duration and string.format('%.0f', meta.eval_duration / 1e6) or 'nil'
+          )
+        end
+        log.error(provider .. ' chat request failed: ' .. error_message .. details)
+        vim.notify(provider .. ' request failed: ' .. error_message .. '. See ai-commit logs.', vim.log.levels.ERROR)
         callback(nil, nil)
         return
       end
@@ -260,7 +315,7 @@ local function complete_ollama(full_prompt, callback, status_callback, request_c
       local cleaned = clean_message(message)
       local used_model = meta and meta.used_model or model
       local elapsed_ms = meta and meta.elapsed_ms or 0
-      log.info(string.format('Generated commit message with Ollama (took %.0fms, model=%s)', elapsed_ms, used_model))
+      log.info(string.format('Generated commit message with %s (took %.0fms, model=%s)', provider, elapsed_ms, used_model))
       callback(cleaned, { requested_model = model, used_model = used_model })
     end,
   })
@@ -804,17 +859,18 @@ local function generate_commit_message_async(branch, recent_commits, diff, callb
   log.debug(string.format('Prompt built (branch=%s, commits=%d chars, diff=%d chars)', branch, #recent_commits, #diff))
 
   local ai_provider = require 'ai-provider'
+  local local_provider = CONFIG.local_provider or 'ollama'
   if status_callback then
-    status_callback 'Checking Ollama'
+    status_callback('Checking ' .. local_provider)
   end
 
-  ai_provider.check('ollama', function(working)
+  ai_provider.check(local_provider, function(working)
     if request_context and request_context.is_cancelled and request_context.is_cancelled() then
       return
     end
 
     if working then
-      log.info 'Routing commit message generation to Ollama'
+      log.info('Routing commit message generation to ' .. local_provider)
       complete_ollama(full_prompt, callback, status_callback, request_context)
       return
     end
@@ -871,6 +927,26 @@ local function get_copilot_models()
   return fallback_models
 end
 
+local function select_local_model()
+  local log = setup_logger()
+  local ai_provider = require 'ai-provider'
+
+  ai_provider.select_helper({
+    prompt = 'Select AI commit model:',
+    current = { provider = CONFIG.local_provider, model = CONFIG.local_model },
+  }, function(choice)
+    if not choice then
+      return
+    end
+
+    CONFIG.local_provider = choice.provider
+    CONFIG.local_model = choice.model
+    CONFIG.local_model_name = choice.model
+    log.info('AI commit local model changed to: ' .. choice.label)
+    save_preferences()
+  end)
+end
+
 --- Select and save a model from multiple providers
 local function select_model()
   local log = setup_logger()
@@ -1320,6 +1396,16 @@ end
 return {
   'nvim-lua/plenary.nvim',
   ft = 'gitcommit',
+  cmd = { 'AICommit', 'AICommitModel' },
+  keys = {
+    {
+      '<leader>psc',
+      function()
+        select_local_model()
+      end,
+      desc = 'AI [P]rovider [S]elect [C]ommit',
+    },
+  },
 
   config = function()
     -- Load saved model preference
@@ -1349,5 +1435,9 @@ return {
       desc = 'Generate AI commit message',
     })
 
+    vim.api.nvim_create_user_command('AICommitModel', select_local_model, {
+      desc = 'Select AI commit model',
+    })
+
   end,
 }
diff --git a/lua/plugins/ai-provider.lua b/lua/plugins/ai-provider.lua
index 4b3825e..9a96306 100644
--- a/lua/plugins/ai-provider.lua
+++ b/lua/plugins/ai-provider.lua
@@ -11,6 +11,7 @@ return {
       ollama = {
         default_model = 'gemma4:e2b',
         context_size = 1024 * 8,
+        load_timeout = 120000,
         keep_alive = '4h',
         models = {
           ['gemma4:e2b 32k'] = {
@@ -21,6 +22,15 @@ return {
             model = 'gemma4:e2b',
             context_size = 1024 * 64,
           },
+          ['qwen3.5:4b 128k'] = {
+            model = 'qwen3.5:4b',
+            context_size = 1024 * 128,
+          },
+          ['qwen3.5:4b 256k'] = {
+            model = 'qwen3.5:4b',
+            context_size = 1024 * 256,
+            think = false,
+          },
         },
       },
     },