@yanlinglabs/winter-provider-catalog 0.0.20 → 0.0.22

package/overlay/pricing.json

@@ -0,0 +1,243 @@
+{
+  "$comment": "HAND-AUTHORED provider/model price evidence. This is intentionally a narrow patch layer rather than a model overlay: replacing an extracted model row merely to add a price would discard its endpoint, modalities, context window, and effort evidence. All values are USD per million tokens. Where a first-party table is CNY-only, its sourceRef records the exact selected CNY tier and the ECB reference-rate conversion used; `ModelPricing` cannot represent currency, region, or multiple context tiers.",
+  "pricing": {
+    "baidu/ernie-4.5-turbo-128k": {
+      "value": { "inputPerMTokUsd": 0.119445, "outputPerMTokUsd": 0.47778, "cacheReadPerMTokUsd": 0.029861 },
+      "source": "official-doc",
+      "sourceRef": "https://cloud.baidu.com/doc/qianfan-docs/s/Jm8r1826a — ERNIE-4.5-Turbo-128K online list rate: CNY 0.0008 input, CNY 0.0032 output, and CNY 0.0002 cache-hit input per 1K tokens. Converted to USD/MTok using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "baidu/ernie-4.5-turbo-32k": {
+      "value": { "inputPerMTokUsd": 0.119445, "outputPerMTokUsd": 0.47778, "cacheReadPerMTokUsd": 0.029861 },
+      "source": "official-doc",
+      "sourceRef": "https://cloud.baidu.com/doc/qianfan-docs/s/Jm8r1826a — ERNIE-4.5-Turbo-32K online list rate: CNY 0.0008 input, CNY 0.0032 output, and CNY 0.0002 cache-hit input per 1K tokens. Converted to USD/MTok using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "baidu/ernie-4.5-turbo-vl": {
+      "value": { "inputPerMTokUsd": 0.447919, "outputPerMTokUsd": 1.343756 },
+      "source": "official-doc",
+      "sourceRef": "https://cloud.baidu.com/doc/qianfan-docs/s/Jm8r1826a — ERNIE-4.5-Turbo-VL online list rate: CNY 0.003 input and CNY 0.009 output per 1K tokens. Converted to USD/MTok using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "baidu/ernie-5.0": {
+      "value": { "inputPerMTokUsd": 1.493062, "outputPerMTokUsd": 5.972249 },
+      "source": "official-doc",
+      "sourceRef": "https://cloud.baidu.com/doc/qianfan-docs/s/Jm8r1826a — ERNIE-5.0 online list rate is tiered by input length: CNY 0.006/0.024 per 1K tokens at <=32K and CNY 0.01/0.04 above 32K through 128K. This records the higher CNY 0.01 input / CNY 0.04 output tier so estimates do not under-report long prompts. Converted to USD/MTok using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "baidu/ernie-5.1": {
+      "value": { "inputPerMTokUsd": 0.895837, "outputPerMTokUsd": 3.284737 },
+      "source": "official-doc",
+      "sourceRef": "https://cloud.baidu.com/doc/qianfan-docs/s/Jm8r1826a — ERNIE-5.1 online list rate is tiered by input length: CNY 0.004/0.018 per 1K tokens at <=32K and CNY 0.006/0.022 above 32K through 128K. This records the higher CNY 0.006 input / CNY 0.022 output tier so estimates do not under-report long prompts. Converted to USD/MTok using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "qianfan/ernie-5.0-thinking-latest": {
+      "value": { "inputPerMTokUsd": 1.493062, "outputPerMTokUsd": 5.972249 },
+      "source": "official-doc",
+      "sourceRef": "https://cloud.baidu.com/doc/qianfan-docs/s/Jm8r1826a — the ERNIE-5.0 row explicitly includes ERNIE-5.0-Thinking-Latest. Its list rate is CNY 0.006/0.024 per 1K tokens at <=32K and CNY 0.01/0.04 above 32K through 128K. This records the higher CNY 0.01 input / CNY 0.04 output tier. Converted to USD/MTok using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "qianfan/ernie-5.1": {
+      "value": { "inputPerMTokUsd": 0.895837, "outputPerMTokUsd": 3.284737 },
+      "source": "official-doc",
+      "sourceRef": "https://cloud.baidu.com/doc/qianfan-docs/s/Jm8r1826a — ERNIE-5.1 online list rate is CNY 0.004/0.018 per 1K tokens at <=32K and CNY 0.006/0.022 above 32K through 128K. This records the higher CNY 0.006 input / CNY 0.022 output tier. Converted to USD/MTok using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "qwen-cloud/qwen3.5-122b-a10b": {
+      "value": { "inputPerMTokUsd": 0.4, "outputPerMTokUsd": 3.2 },
+      "source": "official-doc",
+      "sourceRef": "https://www.qwencloud.com/models/qwen3.5-122b-a10b — QwenCloud lists $0.40 input and $3.20 output per 1M tokens for this exact API model (retrieved 2026-09-19).",
+      "confidence": "declared",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "qwen-cloud/qwen3.5-397b-a17b": {
+      "value": { "inputPerMTokUsd": 0.6, "outputPerMTokUsd": 3.6 },
+      "source": "official-doc",
+      "sourceRef": "https://www.qwencloud.com/models/qwen3.5-397b-a17b — QwenCloud lists $0.60 input and $3.60 output per 1M tokens for this exact API model (retrieved 2026-09-19).",
+      "confidence": "declared",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "qwen-cloud/qwen3.5-plus-2026-04-20": {
+      "value": { "inputPerMTokUsd": 0.4, "outputPerMTokUsd": 2.4, "cacheReadPerMTokUsd": 0.04, "cacheWritePerMTokUsd": 0.5 },
+      "source": "official-doc",
+      "sourceRef": "https://www.qwencloud.com/models/qwen3.5-plus-2026-04-20 — QwenCloud's <=256K tier lists $0.40 input, $2.40 output, $0.50 explicit cache creation, and $0.04 explicit cache read per 1M tokens. The page also advertises a 1M context, but publishes no rate above 256K; ModelPricing therefore records the published <=256K rate only (retrieved 2026-09-19).",
+      "confidence": "declared",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "qwen-cloud/qwen3.6-27b": {
+      "value": { "inputPerMTokUsd": 0.6, "outputPerMTokUsd": 3.6 },
+      "source": "official-doc",
+      "sourceRef": "https://www.qwencloud.com/models/qwen3.6-27b — QwenCloud lists $0.60 input and $3.60 output per 1M tokens for this exact API model (retrieved 2026-09-19).",
+      "confidence": "declared",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "qwen-cloud/qwen3.6-35b-a3b": {
+      "value": { "inputPerMTokUsd": 0.375, "outputPerMTokUsd": 2.25 },
+      "source": "official-doc",
+      "sourceRef": "https://www.qwencloud.com/models/qwen3.6-35b-a3b — QwenCloud lists $0.375 input and $2.25 output per 1M tokens for this exact API model (retrieved 2026-09-19).",
+      "confidence": "declared",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "qwen-cloud/qwen3.6-plus": {
+      "value": { "inputPerMTokUsd": 0.5, "outputPerMTokUsd": 3, "cacheReadPerMTokUsd": 0.05, "cacheWritePerMTokUsd": 0.625 },
+      "source": "official-doc",
+      "sourceRef": "https://www.qwencloud.com/models/qwen3.6-plus — QwenCloud's <=256K tier lists $0.50 input, $3.00 output, $0.625 explicit cache creation, and $0.05 explicit cache read per 1M tokens. The page also advertises a 1M context, but publishes no rate above 256K; ModelPricing therefore records the published <=256K rate only (retrieved 2026-09-19).",
+      "confidence": "declared",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "qwen-cloud/qwen3.7-max-2026-06-08": {
+      "value": { "inputPerMTokUsd": 2.5, "outputPerMTokUsd": 7.5, "cacheReadPerMTokUsd": 0.5, "cacheWritePerMTokUsd": 3.125 },
+      "source": "official-doc",
+      "sourceRef": "https://www.qwencloud.com/models/qwen3.7-max-2026-06-08 — QwenCloud lists $2.50 input, $7.50 output, $0.50 implicit-cache input, and $3.125 explicit-cache creation per 1M tokens. It separately lists $0.25 explicit cache read; the one cacheRead field records the higher implicit-cache rate so estimates do not under-report either cache mode (retrieved 2026-09-19).",
+      "confidence": "declared",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "qwen-cloud/qwen3.7-plus": {
+      "value": { "inputPerMTokUsd": 0.32, "outputPerMTokUsd": 1.28, "cacheReadPerMTokUsd": 0.064, "cacheWritePerMTokUsd": 0.4 },
+      "source": "official-doc",
+      "sourceRef": "https://www.qwencloud.com/models/qwen3.7-plus — QwenCloud's current <=256K 20%-off price is $0.32 input, $1.28 output, $0.064 implicit-cache input, and $0.40 explicit-cache creation per 1M tokens. It separately lists $0.032 explicit cache read; the one cacheRead field records the higher implicit-cache rate. The undiscounted $0.40/$1.60 list price and the current promotion are both disclosed by the page (retrieved 2026-09-19).",
+      "confidence": "declared",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "qwen-cloud/qwen3.8-max": {
+      "value": { "inputPerMTokUsd": 2, "outputPerMTokUsd": 6, "cacheReadPerMTokUsd": 0.25, "cacheWritePerMTokUsd": 2.5 },
+      "source": "official-doc",
+      "sourceRef": "https://www.qwencloud.com/models/qwen3.8-max — QwenCloud lists $2 input, $6 output, $0.25 implicit-cache input, and $2.50 explicit-cache creation per 1M tokens. It separately lists $0.17 explicit cache read; the one cacheRead field records the higher implicit-cache rate so estimates do not under-report either cache mode (retrieved 2026-09-19).",
+      "confidence": "declared",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba/qwen3.5-122b-a10b": {
+      "value": { "inputPerMTokUsd": 0.438363, "outputPerMTokUsd": 3.506606 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — Model Studio International (Singapore) lists qwen3.5-122b-a10b at CNY 2.936 input / CNY 23.486 output per MTok. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba/qwen3.5-397b-a17b": {
+      "value": { "inputPerMTokUsd": 0.657545, "outputPerMTokUsd": 3.94482 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — Model Studio International (Singapore) lists qwen3.5-397b-a17b at CNY 4.404 input / CNY 26.421 output per MTok. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba/qwen3.5-plus": {
+      "value": { "inputPerMTokUsd": 0.547954, "outputPerMTokUsd": 3.287425 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — Model Studio International (Singapore) lists qwen3.5-plus at CNY 2.936/17.614 per MTok through 256K and CNY 3.67/22.018 at 256K-1M. ModelPricing records the higher CNY 3.67 input / CNY 22.018 output tier so estimates do not under-report long prompts. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba/qwen3.6-27b": {
+      "value": { "inputPerMTokUsd": 0.671358, "outputPerMTokUsd": 4.028151 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — Model Studio International (Singapore) lists qwen3.6-27b at CNY 4.49652 input / CNY 26.97912 output per MTok. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba/qwen3.6-35b-a3b": {
+      "value": { "inputPerMTokUsd": 0.419599, "outputPerMTokUsd": 2.517594 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — Model Studio International (Singapore) lists qwen3.6-35b-a3b at CNY 2.810325 input / CNY 16.86195 output per MTok. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba/qwen3.6-plus": {
+      "value": { "inputPerMTokUsd": 2.237862, "outputPerMTokUsd": 6.713555 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — Model Studio International (Singapore) lists qwen3.6-plus at CNY 3.7471/22.4826 per MTok through 256K and CNY 14.9884/44.965 at 256K-1M. ModelPricing records the higher CNY 14.9884 input / CNY 44.965 output tier so estimates do not under-report long prompts. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba/qwen3.7-max": {
+      "value": { "inputPerMTokUsd": 2.797402, "outputPerMTokUsd": 8.392056 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — Model Studio International (Singapore) lists qwen3.7-max at CNY 18.736 input / CNY 56.207 output per MTok. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba/qwen3.7-plus": {
+      "value": { "inputPerMTokUsd": 1.342711, "outputPerMTokUsd": 5.370844 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — Model Studio International (Singapore) lists qwen3.7-plus at CNY 2.998/11.991 per MTok through 256K and CNY 8.993/35.972 at 256K-1M, marked as a limited-time 20%-off list price. ModelPricing records the higher CNY 8.993 input / CNY 35.972 output list tier so estimates do not under-report long prompts; it may over-report while the promotion applies. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba/qwen3.8-max": {
+      "value": { "inputPerMTokUsd": 2.237802, "outputPerMTokUsd": 6.713555 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — Model Studio International (Singapore) lists qwen3.8-max at CNY 14.988 input / CNY 44.965 output per MTok. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba-cn/qwen3.5-122b-a10b": {
+      "value": { "inputPerMTokUsd": 0.298612, "outputPerMTokUsd": 2.3889 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — China (Beijing) lists qwen3.5-122b-a10b at CNY 0.8/6.4 per MTok through 128K and CNY 2/16 at 128K-256K. ModelPricing records the higher CNY 2 input / CNY 16 output tier. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba-cn/qwen3.5-397b-a17b": {
+      "value": { "inputPerMTokUsd": 0.447919, "outputPerMTokUsd": 2.687512 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — China (Beijing) lists qwen3.5-397b-a17b at CNY 1.2/7.2 per MTok through 128K and CNY 3/18 at 128K-256K. ModelPricing records the higher CNY 3 input / CNY 18 output tier. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba-cn/qwen3.5-plus": {
+      "value": { "inputPerMTokUsd": 0.597225, "outputPerMTokUsd": 3.58335 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — China (Beijing) lists qwen3.5-plus at CNY 0.8/4.8 per MTok through 128K, CNY 2/12 through 256K, and CNY 4/24 at 256K-1M. ModelPricing records the higher CNY 4 input / CNY 24 output tier. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba-cn/qwen3.6-27b": {
+      "value": { "inputPerMTokUsd": 0.447919, "outputPerMTokUsd": 2.687512 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — China (Beijing) lists qwen3.6-27b at CNY 3 input / CNY 18 output per MTok. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba-cn/qwen3.6-35b-a3b": {
+      "value": { "inputPerMTokUsd": 0.268751, "outputPerMTokUsd": 1.612507 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — China (Beijing) lists qwen3.6-35b-a3b at CNY 1.8 input / CNY 10.8 output per MTok. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba-cn/qwen3.6-plus": {
+      "value": { "inputPerMTokUsd": 1.19445, "outputPerMTokUsd": 7.166699 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — China (Beijing) lists qwen3.6-plus at CNY 2/12 per MTok through 256K and CNY 8/48 at 256K-1M. ModelPricing records the higher CNY 8 input / CNY 48 output tier. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba-cn/qwen3.7-max": {
+      "value": { "inputPerMTokUsd": 1.791675, "outputPerMTokUsd": 5.375024 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — China (Beijing) lists qwen3.7-max at CNY 12 input / CNY 36 output per MTok. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba-cn/qwen3.7-plus": {
+      "value": { "inputPerMTokUsd": 0.895837, "outputPerMTokUsd": 3.58335 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — China (Beijing) lists qwen3.7-plus at list CNY 2/8 per MTok through 256K and CNY 6/24 at 256K-1M, both marked as limited-time 20%-off. ModelPricing records the higher CNY 6 input / CNY 24 output list tier so estimates do not under-report long prompts; it may over-report while the promotion applies. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    },
+    "alibaba-cn/qwen3.8-max": {
+      "value": { "inputPerMTokUsd": 1.791675, "outputPerMTokUsd": 5.375024 },
+      "source": "official-doc",
+      "sourceRef": "https://help.aliyun.com/en/model-studio/model-pricing — China (Beijing) lists qwen3.8-max at CNY 12 input / CNY 36 output per MTok. Converted using the ECB 2026-09-18 reference rates (EUR 1 = USD 1.1460; CNY 7.6755): https://www.ecb.europa.eu/stats/shared/pdf/eurofxref.pdf?57eae2dc36bcb4af234d5e0b7e1a28c3= (retrieved 2026-09-19).",
+      "confidence": "inferred",
+      "observedAt": "2026-09-19T00:00:00Z"
+    }
+  }
+}

package/overlay/providers.json

@@ -884,6 +884,42 @@
       },
       "scope": "llm"
     },
+    {
+      "id": "kimi-coding-openai",
+      "displayName": "Kimi Code (OpenAI dialect)",
+      "$comment": "The Kimi Code membership endpoint supports both documented dialects. This separate subscription row is the OpenAI-compatible sibling of `kimi-coding`, not Moonshot's token-billed platform API.",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.kimi.com/coding/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Membership API-key access, plan-dependent model entitlements, and the OpenAI-compatible endpoint are documented, but no Winter live gate has exercised this subscription surface."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://www.kimi.com/code/docs/en/kimi-code/models.html — Kimi Code's current model configuration documents the OpenAI-compatible Coding endpoint at https://api.kimi.com/coding/v1, membership API keys, and third-party tool use (retrieved 2026-09-19).",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
     {
       "id": "lemonade",
       "displayName": "Lemonade Server (local)",
@@ -2273,16 +2309,16 @@
     },
     {
       "id": "zai",
-      "displayName": "Z.AI GLM Coding (OpenAI dialect)",
+      "displayName": "Z.AI API (OpenAI dialect)",
       "protocols": [
         "openai-chat-completions"
       ],
-      "$comment": "R6b-5 (WS-13b §2): a DIALECT SIBLING. The vendor documents two wire dialects at two different base URLs, and a provider row carries exactly one `adapterId`, so two dialects are two rows — each with its own `defaultEndpoints.api`, its own model keys, and the dialect in its `displayName`. `winter.anthropic-messages` is multi-provider in fact, measured rather than assumed: `runtime/src/provider/catalog-endpoint-shape.test.ts` drives two sibling rows against a loopback fake and each reaches ITS OWN `<root>/v1/messages`. A WINTER-OWNED row (`upstream.project: \"winter\"`, empty commit): it is not extracted, so nothing here came from the pinned tree's executable half. The endpoint and dialect are the VENDOR'S OWN documented values, fetched and read on 2026-09-06 — the `fetched-document` citation tier (PROVENANCE.md). The OPENAI half of z.ai's GLM Coding Plan. Its Anthropic sibling is the EXTRACTED `zai-anthropic` row — upstream's own `zai` entry is `format: \"claude\"` at `https://api.z.ai/api/anthropic/v1/messages`, which z.ai's Claude-client doc confirms verbatim, so the allowlist admits that id under the sibling id rather than under `zai` (the `gemini` -> `google` precedent). Upstream's `glm`/`glm-cn`/`glmt` ids reach this same endpoint through a bespoke `glm` executor Winter does not represent; they are in `blocked` pointing here.",
+      "$comment": "Normal Z.AI token-billed API. This is deliberately distinct from `zai-coding`, whose Coding Plan key uses the different `/api/coding/paas/v4` endpoint and spends subscription credits rather than API balance. The provider documentation shows this ordinary OpenAI-compatible endpoint in its GLM-5.2 quick start; its pricing table is the basis for the token rates on this provider's models. The Anthropic-compatible route remains separately represented as `zai-anthropic` because adapters are dialect-specific.",
       "authKinds": [
         "api-key"
       ],
       "defaultEndpoints": {
-        "api": "https://api.z.ai/api/coding/paas/v4"
+        "api": "https://api.z.ai/api/paas/v4"
       },
       "modelDiscovery": "openai-models",
       "liveCatalogAuthority": "unknown",
@@ -2296,13 +2332,87 @@
       "risk": {
         "class": "review-required",
         "reasons": [
-          "admission is fetched-document, but no live gate case has run against this endpoint; the model roster is the sibling's and is unverified on this dialect"
+          "admission is fetched-document, but no live gate case has run against this endpoint"
         ]
       },
       "pricingBasis": "token",
       "admission": {
         "basis": "api-key",
-        "citation": "https://docs.z.ai/devpack/tool/others — z.ai's own GLM Coding Plan docs, retrieved 2026-09-06: the \"Coding Endpoint\" table lists `https://api.z.ai/api/coding/paas/v4` under \"OpenAI Chat Completions\", and the Cline configuration example repeats it.",
+        "citation": "https://docs.z.ai/guides/llm/glm-5.2 — Z.AI's own normal API quick start, retrieved 2026-09-19: the OpenAI-compatible client uses an API key and base URL `https://api.z.ai/api/paas/v4/`; the separately documented Coding Plan endpoint is not used here.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "zai-coding",
+      "displayName": "Z.AI GLM Coding Plan (OpenAI dialect)",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "$comment": "The GLM Coding Plan is a distinct subscription entitlement, not the normal Z.AI API balance. It uses a Coding Plan API key and `https://api.z.ai/api/coding/paas/v4`; model usage consumes plan credits subject to five-hour and weekly limits. `pricingBasis: subscription` intentionally keeps every model's token price undefined: ModelPricing cannot represent the plan's credits, rolling quotas, peak/off-peak discount, or plan tier. The provider's current plan documentation lists GLM-5.3 and GLM-5.3-Flash as the effective models; documented legacy ids are aliases on those two rows rather than separately priced models.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.z.ai/api/coding/paas/v4"
+      },
+      "modelDiscovery": "openai-models",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "The vendor limits subscription benefits to its listed tools and environments; although its current list names Codex, this application has not run a live plan-key gate against the endpoint.",
+          "Plan entitlement, context, and rate limits vary by current vendor plan and are not discoverable from the normal API's `/models` response."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://docs.z.ai/devpack/overview and https://docs.z.ai/devpack/tool/others — Z.AI's current GLM Coding Plan documentation, retrieved 2026-09-19: it is a subscription package; the OpenAI Chat Completions Coding Endpoint is `https://api.z.ai/api/coding/paas/v4`; supported tools include Codex; and plan usage is governed by five-hour and weekly credit limits.",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "zai-coding-anthropic",
+      "displayName": "Z.AI GLM Coding Plan (Anthropic dialect)",
+      "protocols": [
+        "anthropic-messages"
+      ],
+      "$comment": "Dialect sibling of `zai-coding`. Z.AI's current GLM Coding Plan documentation explicitly supports Anthropic Messages at this root and gives a Claude Code configuration using the same plan API key. It is still subscription-priced: the plan's credits and limits are neither per-token list prices nor representable by ModelPricing.",
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.z.ai/api/anthropic"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.anthropic-messages",
+      "family": "anthropic",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "The vendor documents the Claude Code configuration but this application has not run a live plan-key gate against the Anthropic-compatible endpoint.",
+          "The published OpenAI Chat reasoning replay contract must not be assumed for Anthropic Messages without a dialect-specific document or probe."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://docs.z.ai/devpack/tool/claude and https://docs.z.ai/devpack/tool/others — Z.AI's current GLM Coding Plan documentation, retrieved 2026-09-19: the Claude Code configuration uses `ANTHROPIC_BASE_URL=https://api.z.ai/api/anthropic`; the Coding Plan supports the Anthropic protocol; and the plan is subscription credit based.",
         "tier": "fetched-document"
       },
       "scope": "llm"
@@ -2415,6 +2525,78 @@
         "tier": "audit"
       },
       "scope": "llm"
+    },
+    {
+      "id": "minimax-token-plan",
+      "displayName": "MiniMax Token Plan (OpenAI dialect)",
+      "$comment": "Subscription-only sibling of `minimax`: MiniMax documents a Token Plan Subscription Key that is distinct from a pay-as-you-go API key. The shared OpenAI-compatible endpoint does not make the two billing boundaries interchangeable.",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.minimax.io/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "MiniMax documents plan-key use in OpenAI-compatible tools, but Winter has not run a subscription-key live gate or verified entitlement differences among plan tiers."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://platform.minimax.io/subscribe/coding-plan and https://platform.minimax.io/protocol/paid-agreement — MiniMax's Token Plan page documents use with MiniMax Code and OpenAI-compatible tools; its terms state that a Token Plan Subscription Key and normal pay-as-you-go API Key are separate (retrieved 2026-09-19).",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
+    },
+    {
+      "id": "novita-coding",
+      "displayName": "Novita Coding Plan",
+      "$comment": "Subscription-only sibling of `novita`: the Coding Plan publishes a fixed nine-model allowance on Novita's normal OpenAI-compatible API surface. It has no per-token price fields because plan credits, not ordinary API metering, govern it.",
+      "protocols": [
+        "openai-chat-completions"
+      ],
+      "authKinds": [
+        "api-key"
+      ],
+      "defaultEndpoints": {
+        "api": "https://api.novita.ai/openai/v1"
+      },
+      "modelDiscovery": "none",
+      "liveCatalogAuthority": "unknown",
+      "adapterId": "winter.openai-chat-completions",
+      "family": "openai",
+      "upstream": {
+        "project": "winter",
+        "commit": "",
+        "sourcePaths": []
+      },
+      "risk": {
+        "class": "review-required",
+        "reasons": [
+          "Novita documents the Coding Plan model set and its API-key OpenAI-compatible surface, but Winter has not run a plan-key live gate or established tier-specific rate limits."
+        ]
+      },
+      "pricingBasis": "subscription",
+      "admission": {
+        "basis": "api-key",
+        "citation": "https://novita.ai/coding-plan and https://docs.novita.ai/guides/introduction — Novita's Coding Plan lists its nine included models; Novita's current docs declare Bearer API-key authentication and the OpenAI-compatible base https://api.novita.ai/openai (retrieved 2026-09-19).",
+        "tier": "fetched-document"
+      },
+      "scope": "llm"
     }
   ]
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@yanlinglabs/winter-provider-catalog",
-  "version": "0.0.20",
+  "version": "0.0.22",
   "license": "MIT",
   "type": "module",
   "engines": {