diff --git a/CHANGELOG.md b/CHANGELOG.md
index fc84647..e26e8e5 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,5 +1,8 @@
 ## [Unreleased]
 
+### Fixed
+- `Raix::Configuration` no longer defaults `temperature` to `0.0`. The default was being injected into every request payload, which OpenRouter rejects with `404 No endpoints found that can handle the requested parameters` when routed to providers whose `supported_parameters` list omits `temperature` (notably Anthropic's Claude 4.7 family) and `provider.require_parameters: true` is set — which Raix sets automatically whenever `json: true` is passed to `chat_completion`. Callers that want a specific temperature should set one explicitly (`self.temperature = 0.0` on the including class, or `Raix.configure { |c| c.temperature = 0.0 }` globally); when unset, Raix now omits the parameter and the provider's own server-side default applies. `max_tokens`, `max_completion_tokens`, and `model` defaults are unchanged.
+
 ## [2.0.4] - 2026-05-19
 
 ### Fixed
diff --git a/lib/raix/configuration.rb b/lib/raix/configuration.rb
index bb8859c..bc23405 100644
--- a/lib/raix/configuration.rb
+++ b/lib/raix/configuration.rb
@@ -51,12 +51,19 @@ def self.attr_accessor_with_fallback(method_name)
     DEFAULT_MAX_TOKENS = 1000
     DEFAULT_MAX_COMPLETION_TOKENS = 16_384
     DEFAULT_MODEL = "meta-llama/llama-3.3-8b-instruct:free"
-    DEFAULT_TEMPERATURE = 0.0
     DEFAULT_MAX_TOOL_CALLS = 25
 
     # Initializes a new instance of the Configuration class with default values.
+    #
+    # Note: temperature is intentionally not defaulted. Setting a non-nil
+    # temperature here would force it into every request payload, and some
+    # providers (e.g. Anthropic's Claude 4.7 family on OpenRouter) do not list
+    # `temperature` in their supported parameters. Combined with
+    # `provider.require_parameters: true` (which Raix sets when `json: true`),
+    # an injected default of 0.0 causes OpenRouter to reject the request with
+    # "No endpoints found that can handle the requested parameters." Callers
+    # who want a specific temperature should set one explicitly.
     def initialize(fallback: nil)
-      self.temperature = DEFAULT_TEMPERATURE
       self.max_completion_tokens = DEFAULT_MAX_COMPLETION_TOKENS
       self.max_tokens = DEFAULT_MAX_TOKENS
       self.model = DEFAULT_MODEL
diff --git a/spec/raix/default_temperature_spec.rb b/spec/raix/default_temperature_spec.rb
new file mode 100644
index 0000000..e83ac86
--- /dev/null
+++ b/spec/raix/default_temperature_spec.rb
@@ -0,0 +1,96 @@
+# frozen_string_literal: true
+
+# Regression coverage: Raix must not inject a default `temperature` into the
+# request when the caller hasn't set one.
+#
+# Some OpenRouter-routed providers (notably Anthropic's Claude 4.7 family) do
+# not list `temperature` in their `supported_parameters`. When `json: true`
+# adds `provider.require_parameters: true` to the payload, an unsolicited
+# `temperature: 0.0` from a Raix default causes OpenRouter to return
+# "No endpoints found that can handle the requested parameters" (404).
+class TemperatureDefaultProbe
+  include Raix::ChatCompletion
+end
+
+RSpec.describe Raix::ChatCompletion, "default temperature handling" do
+  let(:instance) { TemperatureDefaultProbe.new }
+
+  let(:fake_response_message) do
+    instance_double(
+      "RubyLLM::Message",
+      content: "ok",
+      tool_calls: nil,
+      tool_call?: false,
+      input_tokens: 1,
+      output_tokens: 1,
+      model_id: "anthropic/claude-opus-4-7",
+      raw: nil
+    )
+  end
+
+  let(:fake_chat) do
+    instance_double(
+      "RubyLLM::Chat",
+      with_instructions: nil,
+      add_message: nil,
+      with_temperature: nil,
+      with_params: nil,
+      with_tool: nil,
+      ask: fake_response_message,
+      complete: fake_response_message
+    )
+  end
+
+  before { allow(RubyLLM).to receive(:chat).and_return(fake_chat) }
+
+  it "does not call with_temperature when no temperature is set anywhere" do
+    instance.send(:ruby_llm_request,
+                  params: {},
+                  model: "anthropic/claude-opus-4-7",
+                  messages: [{ role: "user", content: "hi" }])
+
+    expect(fake_chat).not_to have_received(:with_temperature)
+  end
+
+  it "does not include temperature in additional params when unset" do
+    instance.send(:ruby_llm_request,
+                  params: {},
+                  model: "anthropic/claude-opus-4-7",
+                  messages: [{ role: "user", content: "hi" }])
+
+    expect(fake_chat).not_to have_received(:with_params) { |kwargs|
+      kwargs.key?(:temperature)
+    }
+  end
+
+  it "still forwards an explicitly-set temperature (including 0.0)" do
+    instance.temperature = 0.0
+    instance.chat_completion(messages: [{ user: "hi" }])
+
+    expect(fake_chat).to have_received(:with_temperature).with(0.0)
+  end
+
+  it "forwards a class-level configured temperature" do
+    klass = Class.new do
+      include Raix::ChatCompletion
+      configure { |config| config.temperature = 0.3 }
+    end
+    klass.new.chat_completion(messages: [{ user: "hi" }])
+
+    expect(fake_chat).to have_received(:with_temperature).with(0.3)
+  end
+end
+
+RSpec.describe Raix::Configuration, "temperature default" do
+  it "leaves temperature unset on a fresh configuration" do
+    expect(described_class.new.temperature).to be_nil
+  end
+
+  it "still defaults max_tokens, max_completion_tokens, and model" do
+    config = described_class.new
+
+    expect(config.max_tokens).to eq(Raix::Configuration::DEFAULT_MAX_TOKENS)
+    expect(config.max_completion_tokens).to eq(Raix::Configuration::DEFAULT_MAX_COMPLETION_TOKENS)
+    expect(config.model).to eq(Raix::Configuration::DEFAULT_MODEL)
+  end
+end