From c9b613659c5b03c1cd54f19ef79fb6abce907ab1 Mon Sep 17 00:00:00 2001
From: zyycn <zyycn@users.noreply.github.com>
Date: Fri, 31 Jul 2026 22:43:18 +0800
Subject: [PATCH] fix(openai): refresh GPT pricing from upstream

Align GPT-5.6 Terra/Luna standard, Fast, cache and long-context rates with the current upstream pricing page, and add GPT-4o/4o-mini cached-input rates.

Co-authored-by: Codex <noreply@openai.com>
---
 .../providers/openai/src/transport/usage.rs   | 20 ++++++++------
 .../providers/openai/tests/transport/usage.rs | 27 +++++++++++++++++++
 2 files changed, 39 insertions(+), 8 deletions(-)

diff --git a/backend/crates/providers/openai/src/transport/usage.rs b/backend/crates/providers/openai/src/transport/usage.rs
index ae04a892f..ec2d68078 100644
--- a/backend/crates/providers/openai/src/transport/usage.rs
+++ b/backend/crates/providers/openai/src/transport/usage.rs
@@ -96,25 +96,29 @@ const PRICING_RULES: &[PricingRule] = &[
         model: "gpt-5.6-sol",
         pricing: ModelPricing::new(50_000, 300_000, 5_000)
             .with_cache_write(125)
-            .with_priority(100_000, 600_000, 10_000),
+            .with_priority(100_000, 600_000, 10_000)
+            .with_long_context(100_000, 450_000, 10_000, 0, 0, 0),
     },
     PricingRule {
         model: "gpt-5.6-terra",
-        pricing: ModelPricing::new(25_000, 150_000, 2_500)
+        pricing: ModelPricing::new(20_000, 120_000, 2_000)
             .with_cache_write(125)
-            .with_priority(50_000, 300_000, 5_000),
+            .with_priority(40_000, 240_000, 4_000)
+            .with_long_context(40_000, 180_000, 4_000, 0, 0, 0),
     },
     PricingRule {
         model: "gpt-5.6-luna",
-        pricing: ModelPricing::new(10_000, 60_000, 1_000)
+        pricing: ModelPricing::new(2_000, 12_000, 200)
             .with_cache_write(125)
-            .with_priority(20_000, 120_000, 2_000),
+            .with_priority(4_000, 24_000, 400)
+            .with_long_context(4_000, 18_000, 400, 0, 0, 0),
     },
     PricingRule {
         model: "gpt-5.6",
         pricing: ModelPricing::new(50_000, 300_000, 5_000)
             .with_cache_write(125)
-            .with_priority(100_000, 600_000, 10_000),
+            .with_priority(100_000, 600_000, 10_000)
+            .with_long_context(100_000, 450_000, 10_000, 0, 0, 0),
     },
     PricingRule {
         model: "gpt-5.5-pro",
@@ -162,11 +166,11 @@ const PRICING_RULES: &[PricingRule] = &[
     },
     PricingRule {
         model: "gpt-4o-mini",
-        pricing: ModelPricing::new(1_500, 6_000, 0),
+        pricing: ModelPricing::new(1_500, 6_000, 750),
     },
     PricingRule {
         model: "gpt-4o",
-        pricing: ModelPricing::new(25_000, 100_000, 0),
+        pricing: ModelPricing::new(25_000, 100_000, 12_500),
     },
     PricingRule {
         model: "gpt-4-turbo",
diff --git a/backend/crates/providers/openai/tests/transport/usage.rs b/backend/crates/providers/openai/tests/transport/usage.rs
index 66f2b1947..139b4be99 100644
--- a/backend/crates/providers/openai/tests/transport/usage.rs
+++ b/backend/crates/providers/openai/tests/transport/usage.rs
@@ -30,6 +30,33 @@ fn billing_breakdown_should_preserve_input_output_and_cache_components() {
     assert_eq!(breakdown.multiplier_percent(), 100);
 }
 
+#[test]
+fn billing_breakdown_should_use_latest_gpt_5_6_and_cached_input_prices() {
+    let terra = openai_billing_breakdown("gpt-5.6-terra", 1, 1, 0, 0, None)
+        .expect("gpt-5.6-terra standard pricing");
+    assert_eq!(terra.total_amount().amount().scaled(), 140_000);
+
+    let terra_fast = openai_billing_breakdown("gpt-5.6-terra", 1, 1, 0, 0, Some("fast"))
+        .expect("gpt-5.6-terra fast pricing");
+    assert_eq!(terra_fast.total_amount().amount().scaled(), 280_000);
+
+    let terra_long = openai_billing_breakdown("gpt-5.6-terra", 272_001, 0, 0, 0, None)
+        .expect("gpt-5.6-terra long-context pricing");
+    assert_eq!(terra_long.total_amount().amount().scaled(), 10_880_040_000);
+
+    let luna = openai_billing_breakdown("gpt-5.6-luna", 1, 1, 0, 0, None)
+        .expect("gpt-5.6-luna standard pricing");
+    assert_eq!(luna.total_amount().amount().scaled(), 14_000);
+
+    let gpt_4o =
+        openai_billing_breakdown("gpt-4o", 1, 1, 1, 0, None).expect("gpt-4o cached-input pricing");
+    assert_eq!(gpt_4o.total_amount().amount().scaled(), 112_500);
+
+    let gpt_4o_mini = openai_billing_breakdown("gpt-4o-mini", 1, 1, 1, 0, None)
+        .expect("gpt-4o-mini cached-input pricing");
+    assert_eq!(gpt_4o_mini.total_amount().amount().scaled(), 6_750);
+}
+
 #[test]
 fn billing_breakdown_should_apply_fast_and_flex_tiers_without_guessing_unknown_models() {
     let fast = openai_billing_breakdown("gpt-5.4", 1, 1, 0, 0, Some("fast")).expect("fast pricing");