diff --git a/src-tauri/src/database/mod.rs b/src-tauri/src/database/mod.rs
index 55bb2ea61..dda8da4a5 100644
--- a/src-tauri/src/database/mod.rs
+++ b/src-tauri/src/database/mod.rs
@@ -50,7 +50,7 @@ use std::sync::Mutex;
 
 /// 当前 Schema 版本号
 /// 每次修改表结构时递增，并在 schema.rs 中添加相应的迁移逻辑
-pub(crate) const SCHEMA_VERSION: i32 = 13;
+pub(crate) const SCHEMA_VERSION: i32 = 14;
 
 /// 安全地序列化 JSON，避免 unwrap panic
 pub(crate) fn to_json_string<T: Serialize>(value: &T) -> Result<String, AppError> {
diff --git a/src-tauri/src/database/schema.rs b/src-tauri/src/database/schema.rs
index d477991e4..63a73f801 100644
--- a/src-tauri/src/database/schema.rs
+++ b/src-tauri/src/database/schema.rs
@@ -189,6 +189,7 @@ impl Database {
             pricing_model TEXT,
             input_tokens INTEGER NOT NULL DEFAULT 0, output_tokens INTEGER NOT NULL DEFAULT 0,
             cache_read_tokens INTEGER NOT NULL DEFAULT 0, cache_creation_tokens INTEGER NOT NULL DEFAULT 0,
+            cache_creation_1h_tokens INTEGER NOT NULL DEFAULT 0,
             input_token_semantics INTEGER NOT NULL DEFAULT 0,
             input_cost_usd TEXT NOT NULL DEFAULT '0', output_cost_usd TEXT NOT NULL DEFAULT '0',
             cache_read_cost_usd TEXT NOT NULL DEFAULT '0', cache_creation_cost_usd TEXT NOT NULL DEFAULT '0',
@@ -485,6 +486,11 @@ impl Database {
                         Self::migrate_v12_to_v13(conn)?;
                         Self::set_user_version(conn, 13)?;
                     }
+                    13 => {
+                        log::info!("迁移数据库从 v13 到 v14（记录 1 小时缓存写入 token）");
+                        Self::migrate_v13_to_v14(conn)?;
+                        Self::set_user_version(conn, 14)?;
+                    }
                     _ => {
                         return Err(AppError::Database(format!(
                             "未知的数据库版本 {version}，无法迁移到 {SCHEMA_VERSION}"
@@ -1354,6 +1360,22 @@ impl Database {
         Ok(())
     }
 
+    /// v13 -> v14：持久化 Anthropic 1-hour cache-write bucket。
+    /// 5-minute/unknown writes can be derived from aggregate cache_creation_tokens;
+    /// storing only the premium bucket keeps existing aggregate APIs stable while
+    /// allowing later cost backfills to retain correct TTL pricing.
+    fn migrate_v13_to_v14(conn: &Connection) -> Result<(), AppError> {
+        if Self::table_exists(conn, "proxy_request_logs")? {
+            Self::add_column_if_missing(
+                conn,
+                "proxy_request_logs",
+                "cache_creation_1h_tokens",
+                "INTEGER NOT NULL DEFAULT 0",
+            )?;
+        }
+        Ok(())
+    }
+
     /// 插入默认模型定价数据
     /// 格式: (model_id, display_name, input, output, cache_read, cache_creation)
     /// 注意: model_id 使用短横线格式（如 claude-haiku-4-5），与 API 返回的模型名称标准化后一致
@@ -2713,7 +2735,7 @@ mod tests {
 
         Database::apply_schema_migrations_on_conn(&conn)?;
 
-        assert_eq!(Database::get_user_version(&conn)?, 13);
+        assert_eq!(Database::get_user_version(&conn)?, 14);
         assert!(Database::has_column(
             &conn,
             "proxy_request_logs",
@@ -2724,6 +2746,11 @@ mod tests {
             "usage_daily_rollups",
             "input_token_semantics"
         )?);
+        assert!(Database::has_column(
+            &conn,
+            "proxy_request_logs",
+            "cache_creation_1h_tokens"
+        )?);
         let log_default: i64 = conn.query_row(
             "SELECT dflt_value = '0' FROM pragma_table_info('proxy_request_logs')
              WHERE name = 'input_token_semantics'",
diff --git a/src-tauri/src/proxy/response_processor.rs b/src-tauri/src/proxy/response_processor.rs
index 238e14420..20b4eb039 100644
--- a/src-tauri/src/proxy/response_processor.rs
+++ b/src-tauri/src/proxy/response_processor.rs
@@ -645,12 +645,14 @@ async fn log_usage_internal(
     let request_id = usage.dedup_request_id();
 
     log::debug!(
-        "[{app_type}] 记录请求日志: id={request_id}, provider={provider_id}, model={model}, streaming={is_streaming}, status={status_code}, latency_ms={latency_ms}, first_token_ms={first_token_ms:?}, session={}, input={}, output={}, cache_read={}, cache_creation={}",
+        "[{app_type}] 记录请求日志: id={request_id}, provider={provider_id}, model={model}, streaming={is_streaming}, status={status_code}, latency_ms={latency_ms}, first_token_ms={first_token_ms:?}, session={}, input={}, output={}, cache_read={}, cache_creation={}, cache_creation_5m={}, cache_creation_1h={}",
         session_id.as_deref().unwrap_or("none"),
         usage.input_tokens,
         usage.output_tokens,
         usage.cache_read_tokens,
-        usage.cache_creation_tokens
+        usage.cache_creation_tokens,
+        usage.cache_creation_5m_tokens,
+        usage.cache_creation_1h_tokens
     );
 
     if let Err(e) = logger.log_with_calculation(
@@ -1004,6 +1006,8 @@ mod tests {
             output_tokens: 0,
             cache_read_tokens: 0,
             cache_creation_tokens: 0,
+            cache_creation_5m_tokens: 0,
+            cache_creation_1h_tokens: 0,
             model: None,
             message_id: None,
         };
@@ -1072,6 +1076,8 @@ mod tests {
             output_tokens: 0,
             cache_read_tokens: 0,
             cache_creation_tokens: 0,
+            cache_creation_5m_tokens: 0,
+            cache_creation_1h_tokens: 0,
             model: None,
             message_id: None,
         };
@@ -1154,6 +1160,8 @@ mod tests {
             output_tokens: 0,
             cache_read_tokens: 0,
             cache_creation_tokens: 0,
+            cache_creation_5m_tokens: 0,
+            cache_creation_1h_tokens: 0,
             model: None,
             message_id: None,
         };
diff --git a/src-tauri/src/proxy/usage/calculator.rs b/src-tauri/src/proxy/usage/calculator.rs
index db8def267..4628a6868 100644
--- a/src-tauri/src/proxy/usage/calculator.rs
+++ b/src-tauri/src/proxy/usage/calculator.rs
@@ -94,9 +94,20 @@ impl CostCalculator {
             Decimal::from(usage.output_tokens) * pricing.output_cost_per_million / million;
         let cache_read_cost =
             Decimal::from(usage.cache_read_tokens) * pricing.cache_read_cost_per_million / million;
-        let cache_creation_cost = Decimal::from(usage.cache_creation_tokens)
-            * pricing.cache_creation_cost_per_million
-            / million;
+        let (cache_creation_unspecified, cache_creation_5m, cache_creation_1h) =
+            usage.normalized_cache_creation_buckets();
+        let cache_creation_5m_price = pricing.cache_creation_cost_per_million;
+        // The stored cache-creation price is the existing 5-minute rate. Anthropic
+        // prices 5m writes at 1.25x input and 1h writes at 2x input, therefore the
+        // corresponding 1h price is 8/5 of the configured 5m price. Unknown providers
+        // without TTL details retain the configured legacy rate.
+        let cache_creation_1h_price =
+            cache_creation_5m_price * Decimal::from(8u32) / Decimal::from(5u32);
+        let cache_creation_cost = (Decimal::from(cache_creation_unspecified)
+            + Decimal::from(cache_creation_5m))
+            * cache_creation_5m_price
+            / million
+            + Decimal::from(cache_creation_1h) * cache_creation_1h_price / million;
 
         // 总成本 = 各项基础成本之和 × 倍率
         let base_total = input_cost + output_cost + cache_read_cost + cache_creation_cost;
@@ -159,6 +170,8 @@ mod tests {
             output_tokens: 500,
             cache_read_tokens: 200,
             cache_creation_tokens: 100,
+            cache_creation_5m_tokens: 0,
+            cache_creation_1h_tokens: 0,
             model: None,
             message_id: None,
         };
@@ -191,6 +204,8 @@ mod tests {
             output_tokens: 500,
             cache_read_tokens: 200,
             cache_creation_tokens: 100,
+            cache_creation_5m_tokens: 0,
+            cache_creation_1h_tokens: 0,
             model: None,
             message_id: None,
         };
@@ -211,6 +226,25 @@ mod tests {
         assert_eq!(cost.total_cost, Decimal::from_str("0.010035").unwrap());
     }
 
+    #[test]
+    fn test_one_hour_cache_creation_uses_premium_rate() {
+        let usage = TokenUsage {
+            cache_creation_tokens: 1_000_000,
+            cache_creation_5m_tokens: 250_000,
+            cache_creation_1h_tokens: 750_000,
+            ..TokenUsage::default()
+        };
+        let pricing = ModelPricing::from_strings("3", "15", "0.3", "3.75").unwrap();
+
+        let cost = CostCalculator::calculate(&usage, &pricing, Decimal::ONE);
+
+        // 250k × $3.75/M + 750k × $6/M = $5.4375.
+        assert_eq!(
+            cost.cache_creation_cost,
+            Decimal::from_str("5.4375").unwrap()
+        );
+    }
+
     #[test]
     fn test_cost_multiplier() {
         let usage = TokenUsage {
@@ -218,6 +252,8 @@ mod tests {
             output_tokens: 0,
             cache_read_tokens: 0,
             cache_creation_tokens: 0,
+            cache_creation_5m_tokens: 0,
+            cache_creation_1h_tokens: 0,
             model: None,
             message_id: None,
         };
@@ -240,6 +276,8 @@ mod tests {
             output_tokens: 500,
             cache_read_tokens: 0,
             cache_creation_tokens: 0,
+            cache_creation_5m_tokens: 0,
+            cache_creation_1h_tokens: 0,
             model: None,
             message_id: None,
         };
@@ -257,6 +295,8 @@ mod tests {
             output_tokens: 1,
             cache_read_tokens: 1,
             cache_creation_tokens: 1,
+            cache_creation_5m_tokens: 0,
+            cache_creation_1h_tokens: 0,
             model: None,
             message_id: None,
         };
diff --git a/src-tauri/src/proxy/usage/logger.rs b/src-tauri/src/proxy/usage/logger.rs
index 072dd5ccf..985fad40a 100644
--- a/src-tauri/src/proxy/usage/logger.rs
+++ b/src-tauri/src/proxy/usage/logger.rs
@@ -81,11 +81,12 @@ impl<'a> UsageLogger<'a> {
             "INSERT OR REPLACE INTO proxy_request_logs (
                 request_id, provider_id, app_type, model, request_model, pricing_model,
                 input_tokens, output_tokens, cache_read_tokens, cache_creation_tokens,
+                cache_creation_1h_tokens,
                 input_token_semantics,
                 input_cost_usd, output_cost_usd, cache_read_cost_usd, cache_creation_cost_usd, total_cost_usd,
                 latency_ms, first_token_ms, status_code, error_message, session_id,
                 provider_type, is_streaming, cost_multiplier, created_at
-            ) VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8, ?9, ?10, ?11, ?12, ?13, ?14, ?15, ?16, ?17, ?18, ?19, ?20, ?21, ?22, ?23, ?24, ?25)",
+            ) VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8, ?9, ?10, ?11, ?12, ?13, ?14, ?15, ?16, ?17, ?18, ?19, ?20, ?21, ?22, ?23, ?24, ?25, ?26)",
             rusqlite::params![
                 log.request_id,
                 log.provider_id,
@@ -97,6 +98,7 @@ impl<'a> UsageLogger<'a> {
                 log.usage.output_tokens,
                 log.usage.cache_read_tokens,
                 log.usage.cache_creation_tokens,
+                log.usage.cache_creation_1h_tokens,
                 input_token_semantics,
                 input_cost,
                 output_cost,
@@ -396,6 +398,8 @@ mod tests {
             output_tokens: 500,
             cache_read_tokens: 0,
             cache_creation_tokens: 0,
+            cache_creation_5m_tokens: 0,
+            cache_creation_1h_tokens: 0,
             model: None,
             message_id: None,
         };
diff --git a/src-tauri/src/proxy/usage/parser.rs b/src-tauri/src/proxy/usage/parser.rs
index 45120b8ae..c28edf8c5 100644
--- a/src-tauri/src/proxy/usage/parser.rs
+++ b/src-tauri/src/proxy/usage/parser.rs
@@ -27,6 +27,22 @@ fn openai_cache_write_tokens(usage: &Value) -> u32 {
         .unwrap_or(0) as u32
 }
 
+fn cache_creation_ttl_tokens(usage: &Value) -> (u32, u32) {
+    let details = usage
+        .get("cache_creation")
+        .or_else(|| usage.pointer("/input_tokens_details/cache_creation"))
+        .or_else(|| usage.pointer("/prompt_tokens_details/cache_creation"));
+    let five_minutes = details
+        .and_then(|value| value.get("ephemeral_5m_input_tokens"))
+        .and_then(Value::as_u64)
+        .unwrap_or(0) as u32;
+    let one_hour = details
+        .and_then(|value| value.get("ephemeral_1h_input_tokens"))
+        .and_then(Value::as_u64)
+        .unwrap_or(0) as u32;
+    (five_minutes, one_hour)
+}
+
 /// Session 日志 request_id 前缀，与 `session_usage.rs` 中的格式保持一致
 pub const SESSION_REQUEST_ID_PREFIX: &str = "session:";
 
@@ -37,6 +53,13 @@ pub struct TokenUsage {
     pub output_tokens: u32,
     pub cache_read_tokens: u32,
     pub cache_creation_tokens: u32,
+    /// Anthropic cache-write TTL detail. The aggregate above remains the stable
+    /// public/storage metric; these buckets make 1-hour writes billable at 2x input
+    /// instead of the 5-minute 1.25x rate.
+    #[serde(default)]
+    pub cache_creation_5m_tokens: u32,
+    #[serde(default)]
+    pub cache_creation_1h_tokens: u32,
     /// 从响应中提取的实际模型名称（如果可用）
     pub model: Option<String>,
     /// 从响应中提取的消息 ID（用于跨源去重）
@@ -47,6 +70,23 @@ pub struct TokenUsage {
 }
 
 impl TokenUsage {
+    /// Return mutually exclusive cache-write buckets, clamped to the aggregate.
+    /// Providers that do not expose TTL detail remain in `unspecified` and keep the
+    /// legacy cache-creation price.
+    pub fn normalized_cache_creation_buckets(&self) -> (u32, u32, u32) {
+        let one_hour = self
+            .cache_creation_1h_tokens
+            .min(self.cache_creation_tokens);
+        let five_minutes = self
+            .cache_creation_5m_tokens
+            .min(self.cache_creation_tokens.saturating_sub(one_hour));
+        let unspecified = self
+            .cache_creation_tokens
+            .saturating_sub(one_hour)
+            .saturating_sub(five_minutes);
+        (unspecified, five_minutes, one_hour)
+    }
+
     /// 生成与 session 日志共享的 request_id，用于跨源去重。
     /// 有 message_id 时返回 `session:{id}`，否则回退到随机 UUID。
     pub fn dedup_request_id(&self) -> String {
@@ -83,6 +123,7 @@ impl TokenUsage {
     /// 从 Claude API 非流式响应解析
     pub fn from_claude_response(body: &Value) -> Option<Self> {
         let usage = body.get("usage")?;
+        let (cache_creation_5m_tokens, cache_creation_1h_tokens) = cache_creation_ttl_tokens(usage);
         // 提取响应中的模型名称
         let model = body
             .get("model")
@@ -104,6 +145,8 @@ impl TokenUsage {
                 .get("cache_creation_input_tokens")
                 .and_then(|v| v.as_u64())
                 .unwrap_or(0) as u32,
+            cache_creation_5m_tokens,
+            cache_creation_1h_tokens,
             model,
             message_id,
         })
@@ -134,6 +177,8 @@ impl TokenUsage {
                             }
                         }
                         if let Some(msg_usage) = event.get("message").and_then(|m| m.get("usage")) {
+                            let (cache_creation_5m, cache_creation_1h) =
+                                cache_creation_ttl_tokens(msg_usage);
                             // 从 message_start 获取 input_tokens（原生 Claude API）
                             if let Some(input) =
                                 msg_usage.get("input_tokens").and_then(|v| v.as_u64())
@@ -150,10 +195,14 @@ impl TokenUsage {
                                 .and_then(|v| v.as_u64())
                                 .unwrap_or(0)
                                 as u32;
+                            usage.cache_creation_5m_tokens = cache_creation_5m;
+                            usage.cache_creation_1h_tokens = cache_creation_1h;
                         }
                     }
                     "message_delta" => {
                         if let Some(delta_usage) = event.get("usage") {
+                            let (delta_cache_creation_5m, delta_cache_creation_1h) =
+                                cache_creation_ttl_tokens(delta_usage);
                             // 从 message_delta 获取 output_tokens
                             if let Some(output) =
                                 delta_usage.get("output_tokens").and_then(|v| v.as_u64())
@@ -192,6 +241,14 @@ impl TokenUsage {
                                     }
                                     if let Some(cache_creation) = delta_cache_creation {
                                         usage.cache_creation_tokens = cache_creation;
+                                        if delta_cache_creation_5m > 0
+                                            || delta_cache_creation_1h > 0
+                                        {
+                                            usage.cache_creation_5m_tokens =
+                                                delta_cache_creation_5m;
+                                            usage.cache_creation_1h_tokens =
+                                                delta_cache_creation_1h;
+                                        }
                                     }
                                 }
                             }
@@ -206,6 +263,10 @@ impl TokenUsage {
                             if usage.cache_creation_tokens == 0 {
                                 if let Some(cache_creation) = delta_cache_creation {
                                     usage.cache_creation_tokens = cache_creation;
+                                    if delta_cache_creation_5m > 0 || delta_cache_creation_1h > 0 {
+                                        usage.cache_creation_5m_tokens = delta_cache_creation_5m;
+                                        usage.cache_creation_1h_tokens = delta_cache_creation_1h;
+                                    }
                                 }
                             }
                         }
@@ -237,6 +298,8 @@ impl TokenUsage {
             output_tokens: usage.get("completion_tokens")?.as_u64()? as u32,
             cache_read_tokens: 0,
             cache_creation_tokens: 0,
+            cache_creation_5m_tokens: 0,
+            cache_creation_1h_tokens: 0,
             model: None,
             message_id: None,
         })
@@ -270,12 +333,15 @@ impl TokenUsage {
 
         let cached_tokens = openai_cache_read_tokens(usage);
         let cache_write_tokens = openai_cache_write_tokens(usage);
+        let (cache_creation_5m_tokens, cache_creation_1h_tokens) = cache_creation_ttl_tokens(usage);
 
         Some(Self {
             input_tokens: input_tokens? as u32,
             output_tokens: output_tokens? as u32,
             cache_read_tokens: cached_tokens,
             cache_creation_tokens: cache_write_tokens,
+            cache_creation_5m_tokens,
+            cache_creation_1h_tokens,
             model,
             message_id: None,
         })
@@ -294,6 +360,7 @@ impl TokenUsage {
         // 获取 cached_tokens (可能在 cache_read_input_tokens 或 input_tokens_details 中)
         let cached_tokens = openai_cache_read_tokens(usage);
         let cache_write_tokens = openai_cache_write_tokens(usage);
+        let (cache_creation_5m_tokens, cache_creation_1h_tokens) = cache_creation_ttl_tokens(usage);
 
         // 调整 input_tokens: OpenAI total input 同时包含 cache read/write 两桶。
         let adjusted_input = input_tokens
@@ -311,6 +378,8 @@ impl TokenUsage {
             output_tokens,
             cache_read_tokens: cached_tokens,
             cache_creation_tokens: cache_write_tokens,
+            cache_creation_5m_tokens,
+            cache_creation_1h_tokens,
             model,
             message_id: None,
         })
@@ -391,6 +460,7 @@ impl TokenUsage {
         // 获取 cached_tokens (可能在 prompt_tokens_details 中)
         let cached_tokens = openai_cache_read_tokens(usage);
         let cache_write_tokens = openai_cache_write_tokens(usage);
+        let (cache_creation_5m_tokens, cache_creation_1h_tokens) = cache_creation_ttl_tokens(usage);
 
         // 提取响应中的模型名称
         let model = body
@@ -403,6 +473,8 @@ impl TokenUsage {
             output_tokens: completion_tokens as u32,
             cache_read_tokens: cached_tokens,
             cache_creation_tokens: cache_write_tokens,
+            cache_creation_5m_tokens,
+            cache_creation_1h_tokens,
             model,
             message_id: None,
         })
@@ -448,6 +520,8 @@ impl TokenUsage {
                 .and_then(|v| v.as_u64())
                 .unwrap_or(0) as u32,
             cache_creation_tokens: 0,
+            cache_creation_5m_tokens: 0,
+            cache_creation_1h_tokens: 0,
             model,
             message_id: None,
         })
@@ -499,6 +573,8 @@ impl TokenUsage {
                 output_tokens: total_output,
                 cache_read_tokens: total_cache_read,
                 cache_creation_tokens: 0,
+                cache_creation_5m_tokens: 0,
+                cache_creation_1h_tokens: 0,
                 model,
                 message_id: None,
             })
@@ -521,7 +597,11 @@ mod tests {
                 "input_tokens": 100,
                 "output_tokens": 50,
                 "cache_read_input_tokens": 20,
-                "cache_creation_input_tokens": 10
+                "cache_creation_input_tokens": 10,
+                "cache_creation": {
+                    "ephemeral_5m_input_tokens": 4,
+                    "ephemeral_1h_input_tokens": 6
+                }
             }
         });
 
@@ -530,6 +610,9 @@ mod tests {
         assert_eq!(usage.output_tokens, 50);
         assert_eq!(usage.cache_read_tokens, 20);
         assert_eq!(usage.cache_creation_tokens, 10);
+        assert_eq!(usage.cache_creation_5m_tokens, 4);
+        assert_eq!(usage.cache_creation_1h_tokens, 6);
+        assert_eq!(usage.normalized_cache_creation_buckets(), (0, 4, 6));
         assert_eq!(usage.model, Some("claude-sonnet-4-20250514".to_string()));
     }
 
diff --git a/src-tauri/src/services/session_usage.rs b/src-tauri/src/services/session_usage.rs
index 7d8756722..3c5c83660 100644
--- a/src-tauri/src/services/session_usage.rs
+++ b/src-tauri/src/services/session_usage.rs
@@ -448,6 +448,8 @@ fn insert_session_log_entry(
         output_tokens: msg.output_tokens,
         cache_read_tokens: msg.cache_read_tokens,
         cache_creation_tokens: msg.cache_creation_tokens,
+        cache_creation_5m_tokens: 0,
+        cache_creation_1h_tokens: 0,
         model: Some(msg.model.clone()),
         message_id: None,
     };
diff --git a/src-tauri/src/services/session_usage_codex.rs b/src-tauri/src/services/session_usage_codex.rs
index 5757f9ded..7bf33257f 100644
--- a/src-tauri/src/services/session_usage_codex.rs
+++ b/src-tauri/src/services/session_usage_codex.rs
@@ -634,6 +634,8 @@ fn insert_codex_session_entry(
         output_tokens: delta.output,
         cache_read_tokens: delta.cached_input,
         cache_creation_tokens: 0,
+        cache_creation_5m_tokens: 0,
+        cache_creation_1h_tokens: 0,
         model: Some(model.to_string()),
         message_id: None,
     };
diff --git a/src-tauri/src/services/session_usage_gemini.rs b/src-tauri/src/services/session_usage_gemini.rs
index c904726f2..3e81a3613 100644
--- a/src-tauri/src/services/session_usage_gemini.rs
+++ b/src-tauri/src/services/session_usage_gemini.rs
@@ -270,6 +270,8 @@ fn insert_gemini_session_entry(
         output_tokens,
         cache_read_tokens: tokens.cached,
         cache_creation_tokens: 0,
+        cache_creation_5m_tokens: 0,
+        cache_creation_1h_tokens: 0,
         model: Some(model.to_string()),
         message_id: None,
     };
diff --git a/src-tauri/src/services/session_usage_opencode.rs b/src-tauri/src/services/session_usage_opencode.rs
index 83574bb62..f4ec116ab 100644
--- a/src-tauri/src/services/session_usage_opencode.rs
+++ b/src-tauri/src/services/session_usage_opencode.rs
@@ -362,6 +362,8 @@ fn insert_opencode_message(
                 output_tokens: output_with_reasoning,
                 cache_read_tokens: msg.cache_read_tokens,
                 cache_creation_tokens: msg.cache_write_tokens,
+                cache_creation_5m_tokens: 0,
+                cache_creation_1h_tokens: 0,
                 model: Some(msg.model_id.clone()),
                 message_id: None,
             };
diff --git a/src-tauri/src/services/usage_stats.rs b/src-tauri/src/services/usage_stats.rs
index d0760f2c4..0fb576b84 100644
--- a/src-tauri/src/services/usage_stats.rs
+++ b/src-tauri/src/services/usage_stats.rs
@@ -137,6 +137,9 @@ pub struct RequestLogDetail {
     pub output_tokens: u32,
     pub cache_read_tokens: u32,
     pub cache_creation_tokens: u32,
+    /// Internal TTL pricing bucket; aggregate UI metrics remain unchanged.
+    #[serde(skip)]
+    pub cache_creation_1h_tokens: u32,
     /// Internal storage semantics; omitted from the UI/API payload.
     #[serde(skip)]
     pub input_token_semantics: i64,
@@ -159,12 +162,12 @@ pub struct RequestLogDetail {
     pub pricing_model: Option<String>,
 }
 
-/// 把 26 列的查询结果映射为 `RequestLogDetail`。
+/// 把 27 列的查询结果映射为 `RequestLogDetail`。
 ///
-/// 调用方的 SELECT **必须**按以下顺序返回 26 列：
+/// 调用方的 SELECT **必须**按以下顺序返回 27 列：
 /// `request_id, provider_id, provider_name, app_type, model, request_model,
 ///  cost_multiplier, input_tokens, output_tokens, cache_read_tokens,
-///  cache_creation_tokens, input_cost_usd, output_cost_usd, cache_read_cost_usd,
+///  cache_creation_tokens, cache_creation_1h_tokens, input_cost_usd, output_cost_usd, cache_read_cost_usd,
 ///  cache_creation_cost_usd, total_cost_usd, is_streaming, latency_ms,
 ///  first_token_ms, duration_ms, status_code, error_message, created_at,
 ///  data_source, pricing_model, input_token_semantics`
@@ -185,21 +188,22 @@ fn row_to_request_log_detail(row: &rusqlite::Row<'_>) -> rusqlite::Result<Reques
         output_tokens: row.get::<_, i64>(8)? as u32,
         cache_read_tokens: row.get::<_, i64>(9)? as u32,
         cache_creation_tokens: row.get::<_, i64>(10)? as u32,
-        input_cost_usd: row.get(11)?,
-        output_cost_usd: row.get(12)?,
-        cache_read_cost_usd: row.get(13)?,
-        cache_creation_cost_usd: row.get(14)?,
-        total_cost_usd: row.get(15)?,
-        is_streaming: row.get::<_, i64>(16)? != 0,
-        latency_ms: row.get::<_, i64>(17)? as u64,
-        first_token_ms: row.get::<_, Option<i64>>(18)?.map(|v| v as u64),
-        duration_ms: row.get::<_, Option<i64>>(19)?.map(|v| v as u64),
-        status_code: row.get::<_, i64>(20)? as u16,
-        error_message: row.get(21)?,
-        created_at: row.get(22)?,
-        data_source: row.get(23)?,
-        pricing_model: row.get(24)?,
-        input_token_semantics: row.get::<_, i64>(25)?,
+        cache_creation_1h_tokens: row.get::<_, i64>(11)? as u32,
+        input_cost_usd: row.get(12)?,
+        output_cost_usd: row.get(13)?,
+        cache_read_cost_usd: row.get(14)?,
+        cache_creation_cost_usd: row.get(15)?,
+        total_cost_usd: row.get(16)?,
+        is_streaming: row.get::<_, i64>(17)? != 0,
+        latency_ms: row.get::<_, i64>(18)? as u64,
+        first_token_ms: row.get::<_, Option<i64>>(19)?.map(|v| v as u64),
+        duration_ms: row.get::<_, Option<i64>>(20)?.map(|v| v as u64),
+        status_code: row.get::<_, i64>(21)? as u16,
+        error_message: row.get(22)?,
+        created_at: row.get(23)?,
+        data_source: row.get(24)?,
+        pricing_model: row.get(25)?,
+        input_token_semantics: row.get::<_, i64>(26)?,
     })
 }
 
@@ -1530,6 +1534,7 @@ impl Database {
             "SELECT l.request_id, l.provider_id, {logs_pname} as provider_name, l.app_type, l.model,
                     l.request_model, l.cost_multiplier,
                     l.input_tokens, l.output_tokens, l.cache_read_tokens, l.cache_creation_tokens,
+                    l.cache_creation_1h_tokens,
                     l.input_cost_usd, l.output_cost_usd, l.cache_read_cost_usd, l.cache_creation_cost_usd, l.total_cost_usd,
                     l.is_streaming, l.latency_ms, l.first_token_ms, l.duration_ms,
                     l.status_code, l.error_message, l.created_at, l.data_source, l.pricing_model,
@@ -1574,6 +1579,7 @@ impl Database {
             "SELECT l.request_id, l.provider_id, {detail_pname} as provider_name, l.app_type, l.model,
                     l.request_model, l.cost_multiplier,
                     input_tokens, output_tokens, cache_read_tokens, cache_creation_tokens,
+                    cache_creation_1h_tokens,
                     input_cost_usd, output_cost_usd, cache_read_cost_usd, cache_creation_cost_usd, total_cost_usd,
                     is_streaming, latency_ms, first_token_ms, duration_ms,
                     status_code, error_message, created_at, l.data_source, l.pricing_model,
@@ -1730,6 +1736,7 @@ impl Database {
             "SELECT request_id, provider_id, NULL AS provider_name, app_type, model, request_model,
                         cost_multiplier,
                         input_tokens, output_tokens, cache_read_tokens, cache_creation_tokens,
+                        cache_creation_1h_tokens,
                         input_cost_usd, output_cost_usd, cache_read_cost_usd,
                         cache_creation_cost_usd, total_cost_usd, is_streaming, latency_ms,
                         first_token_ms, duration_ms, status_code, error_message, created_at,
@@ -1836,9 +1843,17 @@ impl Database {
         let cache_read_cost = rust_decimal::Decimal::from(log.cache_read_tokens as u64)
             * pricing.cache_read
             / million;
-        let cache_creation_cost = rust_decimal::Decimal::from(log.cache_creation_tokens as u64)
+        let cache_creation_1h_tokens =
+            log.cache_creation_1h_tokens.min(log.cache_creation_tokens) as u64;
+        let cache_creation_standard_tokens =
+            (log.cache_creation_tokens as u64).saturating_sub(cache_creation_1h_tokens);
+        let cache_creation_1h_price = pricing.cache_creation * rust_decimal::Decimal::from(8u32)
+            / rust_decimal::Decimal::from(5u32);
+        let cache_creation_cost = rust_decimal::Decimal::from(cache_creation_standard_tokens)
             * pricing.cache_creation
-            / million;
+            / million
+            + rust_decimal::Decimal::from(cache_creation_1h_tokens) * cache_creation_1h_price
+                / million;
         // 总成本 = 基础成本之和 × 倍率
         let base_total = input_cost + output_cost + cache_read_cost + cache_creation_cost;
         let total_cost = base_total * multiplier;
@@ -2868,6 +2883,50 @@ mod tests {
         Ok(())
     }
 
+    #[test]
+    fn test_backfill_preserves_one_hour_cache_write_pricing() -> Result<(), AppError> {
+        let db = Database::memory()?;
+
+        {
+            let conn = lock_conn!(db.conn);
+            insert_usage_log(
+                &conn,
+                "claude-cache-one-hour",
+                "claude",
+                "p1",
+                "claude-haiku-4-5",
+                "proxy",
+                1000,
+                0,
+                0,
+                0,
+                1_000_000,
+                200,
+                "0",
+            )?;
+            conn.execute(
+                "UPDATE proxy_request_logs
+                 SET cache_creation_1h_tokens = 1000000
+                 WHERE request_id = 'claude-cache-one-hour'",
+                [],
+            )?;
+        }
+
+        assert_eq!(db.backfill_missing_usage_costs()?, 1);
+
+        let conn = lock_conn!(db.conn);
+        let (cache_creation_cost, total_cost): (String, String) = conn.query_row(
+            "SELECT cache_creation_cost_usd, total_cost_usd
+             FROM proxy_request_logs WHERE request_id = 'claude-cache-one-hour'",
+            [],
+            |row| Ok((row.get(0)?, row.get(1)?)),
+        )?;
+        assert_eq!(cache_creation_cost, "2.000000");
+        assert_eq!(total_cost, "2.000000");
+
+        Ok(())
+    }
+
     #[test]
     fn test_get_usage_summary() -> Result<(), AppError> {
         let db = Database::memory()?;