@molecule/api-resource-ai-models 1.2.6 → 1.3.0

package/README.md

@@ -3,7 +3,7 @@ AUTO-GENERATED — DO NOT EDIT THIS FILE.
 Generated by `mlcl sync-docs` from the package's src/index.ts JSDoc + mlcl/registry.json.
 Edits here are overwritten on the next commit (molecule's pre-commit hook regenerates).
 To change this document, edit the module-level JSDoc in src/index.ts.
-Generated: 2026-08-27T00:37:24.634Z
+Generated: 2026-08-28T19:44:44.140Z
 -->
 
 # @molecule/api-resource-ai-models
@@ -728,6 +728,16 @@ All available AI models, grouped by provider, ordered from most to least capable
 To add or remove a model, edit this array. Both the server-side validation
 and the public discovery endpoint will update automatically.
 
+EVERY entry field that shapes the outbound request (`webSearchToolType`,
+`supportedEffortLevels`/`defaultEffortLevel`/`effortBudgetTokens`,
+`maxOutputTokens`, `regions`) must hold for the model's ACTUAL serving
+host(s) in every region — not just look right in a docs table. The
+pre-commit hook enforces this with a LIVE Synthase-shaped probe of each
+added/changed entry (molecule-dev `verify:model-dispatch --staged-catalog`);
+a value the host rejects fails the whole request for every user
+(glm-5.3-flash 2026-08-28: a `webSearchToolType` both zhipu hosts reject
+made every Synthase turn 400/422 while the model itself worked fine).
+
 Effort is each model's OWN native value — there is no abstract scale (see
 {@link ModelDefinition.supportedEffortLevels}):
 
@@ -841,6 +851,16 @@ Sources (verified 2026-07-28; OpenAI re-verified 2026-07-31 after the
   $1.65/$4.951), so nothing would ever select it — and carrying both would
   put two selectable Alibaba flagships in one family. Revisit only if Alibaba
   publishes it as a distinct first-party DashScope model id.)
+  (re-verified 2026-08-28 on the per-model pages, which publish USD rates per
+  region directly — a better source than the CNY pricing table: qwen3.8-max's
+  Singapore implicit-cache rate is $0.25, NOT the 20%/$0.40 the ZH cache doc's
+  standard table implies. That doc names qwen3.8-max, qwen3.8-flash and
+  qwen3.8-2.4t-a95b as exceptions and defers to the console; the model pages
+  are the console's figures. qwen3.7-max's $0.50 (20%) is confirmed unchanged.
+  qwen3.8-flash (2026-08-26) ADDED: $0.15/$0.47, implicit cache $0.016, 1M ctx
+  / 131,072 out, Image+Text+Video input, tools, hybrid thinking ON by default
+  with thinking_budget. It is cn-region-only — api.deepinfra.com has no
+  Qwen/Qwen3.8-Flash — so ALIBABA_US_MODEL_MAP needs no entry.)
 - Zhipu: https://docs.z.ai/guides/overview/pricing + docs.z.ai/guides/llm/
   glm-5.3 (verified 2026-08-26: glm-5.3 shipped 2026-08-14 and IS in the
   catalog — $1.40/$4.40, cached $0.26, i.e. glm-5.2's card unchanged, 1M ctx,

package/dist/models.d.ts

@@ -14,6 +14,16 @@ import type { ModelDefinition } from './types.js';
  * To add or remove a model, edit this array. Both the server-side validation
  * and the public discovery endpoint will update automatically.
  *
+ * EVERY entry field that shapes the outbound request (`webSearchToolType`,
+ * `supportedEffortLevels`/`defaultEffortLevel`/`effortBudgetTokens`,
+ * `maxOutputTokens`, `regions`) must hold for the model's ACTUAL serving
+ * host(s) in every region — not just look right in a docs table. The
+ * pre-commit hook enforces this with a LIVE Synthase-shaped probe of each
+ * added/changed entry (molecule-dev `verify:model-dispatch --staged-catalog`);
+ * a value the host rejects fails the whole request for every user
+ * (glm-5.3-flash 2026-08-28: a `webSearchToolType` both zhipu hosts reject
+ * made every Synthase turn 400/422 while the model itself worked fine).
+ *
  * Effort is each model's OWN native value — there is no abstract scale (see
  * {@link ModelDefinition.supportedEffortLevels}):
  * - A model driven by a provider-native effort/level param lists its provider
@@ -125,6 +135,16 @@ import type { ModelDefinition } from './types.js';
  *   $1.65/$4.951), so nothing would ever select it — and carrying both would
  *   put two selectable Alibaba flagships in one family. Revisit only if Alibaba
  *   publishes it as a distinct first-party DashScope model id.)
+ *   (re-verified 2026-08-28 on the per-model pages, which publish USD rates per
+ *   region directly — a better source than the CNY pricing table: qwen3.8-max's
+ *   Singapore implicit-cache rate is $0.25, NOT the 20%/$0.40 the ZH cache doc's
+ *   standard table implies. That doc names qwen3.8-max, qwen3.8-flash and
+ *   qwen3.8-2.4t-a95b as exceptions and defers to the console; the model pages
+ *   are the console's figures. qwen3.7-max's $0.50 (20%) is confirmed unchanged.
+ *   qwen3.8-flash (2026-08-26) ADDED: $0.15/$0.47, implicit cache $0.016, 1M ctx
+ *   / 131,072 out, Image+Text+Video input, tools, hybrid thinking ON by default
+ *   with thinking_budget. It is cn-region-only — api.deepinfra.com has no
+ *   Qwen/Qwen3.8-Flash — so ALIBABA_US_MODEL_MAP needs no entry.)
  * - Zhipu: https://docs.z.ai/guides/overview/pricing + docs.z.ai/guides/llm/
  *   glm-5.3 (verified 2026-08-26: glm-5.3 shipped 2026-08-14 and IS in the
  *   catalog — $1.40/$4.40, cached $0.26, i.e. glm-5.2's card unchanged, 1M ctx,

package/dist/models.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"models.d.ts","sourceRoot":"","sources":["../src/models.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AAEH,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,YAAY,CAAA;AAEjD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAyIG;AACH,eAAO,MAAM,MAAM,EAAE,SAAS,eAAe,EAm/CnC,CAAA"}
+{"version":3,"file":"models.d.ts","sourceRoot":"","sources":["../src/models.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AAEH,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,YAAY,CAAA;AAEjD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA6JG;AACH,eAAO,MAAM,MAAM,EAAE,SAAS,eAAe,EAymDnC,CAAA"}

package/dist/models.js

@@ -13,6 +13,16 @@
  * To add or remove a model, edit this array. Both the server-side validation
  * and the public discovery endpoint will update automatically.
  *
+ * EVERY entry field that shapes the outbound request (`webSearchToolType`,
+ * `supportedEffortLevels`/`defaultEffortLevel`/`effortBudgetTokens`,
+ * `maxOutputTokens`, `regions`) must hold for the model's ACTUAL serving
+ * host(s) in every region — not just look right in a docs table. The
+ * pre-commit hook enforces this with a LIVE Synthase-shaped probe of each
+ * added/changed entry (molecule-dev `verify:model-dispatch --staged-catalog`);
+ * a value the host rejects fails the whole request for every user
+ * (glm-5.3-flash 2026-08-28: a `webSearchToolType` both zhipu hosts reject
+ * made every Synthase turn 400/422 while the model itself worked fine).
+ *
  * Effort is each model's OWN native value — there is no abstract scale (see
  * {@link ModelDefinition.supportedEffortLevels}):
  * - A model driven by a provider-native effort/level param lists its provider
@@ -124,6 +134,16 @@
  *   $1.65/$4.951), so nothing would ever select it — and carrying both would
  *   put two selectable Alibaba flagships in one family. Revisit only if Alibaba
  *   publishes it as a distinct first-party DashScope model id.)
+ *   (re-verified 2026-08-28 on the per-model pages, which publish USD rates per
+ *   region directly — a better source than the CNY pricing table: qwen3.8-max's
+ *   Singapore implicit-cache rate is $0.25, NOT the 20%/$0.40 the ZH cache doc's
+ *   standard table implies. That doc names qwen3.8-max, qwen3.8-flash and
+ *   qwen3.8-2.4t-a95b as exceptions and defers to the console; the model pages
+ *   are the console's figures. qwen3.7-max's $0.50 (20%) is confirmed unchanged.
+ *   qwen3.8-flash (2026-08-26) ADDED: $0.15/$0.47, implicit cache $0.016, 1M ctx
+ *   / 131,072 out, Image+Text+Video input, tools, hybrid thinking ON by default
+ *   with thinking_budget. It is cn-region-only — api.deepinfra.com has no
+ *   Qwen/Qwen3.8-Flash — so ALIBABA_US_MODEL_MAP needs no entry.)
  * - Zhipu: https://docs.z.ai/guides/overview/pricing + docs.z.ai/guides/llm/
  *   glm-5.3 (verified 2026-08-26: glm-5.3 shipped 2026-08-14 and IS in the
  *   catalog — $1.40/$4.40, cached $0.26, i.e. glm-5.2's card unchanged, 1M ctx,
@@ -174,7 +194,11 @@ export const MODELS = [
         supportsPromptCaching: true,
         supportsTools: true,
         webSearchToolType: 'web_search_20260209',
-        codeExecutionToolType: 'code_execution_20250825',
+        // Current server-tool type per the API reference (verified 2026-08-28);
+        // 'code_execution_20250825' was the BETA HEADER date (code-execution-2025-08-25),
+        // not a tool type — sending it as one is a 400. Not currently sent by
+        // Synthase (request-shape.ts forwards only webSearchToolType).
+        codeExecutionToolType: 'code_execution_20260521',
         webFetchToolType: 'web_fetch_20260209',
         inputPricePerMTok: 10,
         outputPricePerMTok: 50,
@@ -214,7 +238,11 @@ export const MODELS = [
         supportsPromptCaching: true,
         supportsTools: true,
         webSearchToolType: 'web_search_20260209',
-        codeExecutionToolType: 'code_execution_20250825',
+        // Current server-tool type per the API reference (verified 2026-08-28);
+        // 'code_execution_20250825' was the BETA HEADER date (code-execution-2025-08-25),
+        // not a tool type — sending it as one is a 400. Not currently sent by
+        // Synthase (request-shape.ts forwards only webSearchToolType).
+        codeExecutionToolType: 'code_execution_20260521',
         webFetchToolType: 'web_fetch_20260209',
         // Drop-in successor to Opus 4.8 at identical pricing.
         inputPricePerMTok: 5,
@@ -243,7 +271,11 @@ export const MODELS = [
         supportsPromptCaching: true,
         supportsTools: true,
         webSearchToolType: 'web_search_20260209',
-        codeExecutionToolType: 'code_execution_20250825',
+        // Current server-tool type per the API reference (verified 2026-08-28);
+        // 'code_execution_20250825' was the BETA HEADER date (code-execution-2025-08-25),
+        // not a tool type — sending it as one is a 400. Not currently sent by
+        // Synthase (request-shape.ts forwards only webSearchToolType).
+        codeExecutionToolType: 'code_execution_20260521',
         webFetchToolType: 'web_fetch_20260209',
         inputPricePerMTok: 5,
         outputPricePerMTok: 25,
@@ -276,7 +308,11 @@ export const MODELS = [
         supportsPromptCaching: true,
         supportsTools: true,
         webSearchToolType: 'web_search_20260209',
-        codeExecutionToolType: 'code_execution_20250825',
+        // Current server-tool type per the API reference (verified 2026-08-28);
+        // 'code_execution_20250825' was the BETA HEADER date (code-execution-2025-08-25),
+        // not a tool type — sending it as one is a 400. Not currently sent by
+        // Synthase (request-shape.ts forwards only webSearchToolType).
+        codeExecutionToolType: 'code_execution_20260521',
         webFetchToolType: 'web_fetch_20260209',
         // Standard pricing. Intro pricing ($2/$10) applies through 2026-08-31 —
         // billed here at standard so metering never under-charges; revisit after.
@@ -307,7 +343,11 @@ export const MODELS = [
         supportsPromptCaching: true,
         supportsTools: true,
         webSearchToolType: 'web_search_20260209',
-        codeExecutionToolType: 'code_execution_20250825',
+        // Current server-tool type per the API reference (verified 2026-08-28);
+        // 'code_execution_20250825' was the BETA HEADER date (code-execution-2025-08-25),
+        // not a tool type — sending it as one is a 400. Not currently sent by
+        // Synthase (request-shape.ts forwards only webSearchToolType).
+        codeExecutionToolType: 'code_execution_20260521',
         webFetchToolType: 'web_fetch_20260209',
         inputPricePerMTok: 5,
         outputPricePerMTok: 25,
@@ -342,7 +382,11 @@ export const MODELS = [
         supportsPromptCaching: true,
         supportsTools: true,
         webSearchToolType: 'web_search_20260209',
-        codeExecutionToolType: 'code_execution_20250825',
+        // Current server-tool type per the API reference (verified 2026-08-28);
+        // 'code_execution_20250825' was the BETA HEADER date (code-execution-2025-08-25),
+        // not a tool type — sending it as one is a 400. Not currently sent by
+        // Synthase (request-shape.ts forwards only webSearchToolType).
+        codeExecutionToolType: 'code_execution_20260521',
         webFetchToolType: 'web_fetch_20260209',
         inputPricePerMTok: 5,
         outputPricePerMTok: 25,
@@ -373,7 +417,11 @@ export const MODELS = [
         supportsPromptCaching: true,
         supportsTools: true,
         webSearchToolType: 'web_search_20260209',
-        codeExecutionToolType: 'code_execution_20250825',
+        // Current server-tool type per the API reference (verified 2026-08-28);
+        // 'code_execution_20250825' was the BETA HEADER date (code-execution-2025-08-25),
+        // not a tool type — sending it as one is a 400. Not currently sent by
+        // Synthase (request-shape.ts forwards only webSearchToolType).
+        codeExecutionToolType: 'code_execution_20260521',
         webFetchToolType: 'web_fetch_20260209',
         inputPricePerMTok: 3,
         outputPricePerMTok: 15,
@@ -446,7 +494,11 @@ export const MODELS = [
         supportsTools: true,
         // Tools + ANY reasoning is a 400 on /v1/chat/completions for this family.
         toolsRequireReasoningOff: true,
-        webSearchToolType: 'web_search',
+        // NO webSearchToolType: the OpenAI bond calls /v1/chat/completions, which
+        // has no web_search tool type (it is a Responses-API construct), and the
+        // bond deliberately forwards no server tools. Advertising one here surfaced
+        // web search in the system prompt while it could never work. Re-add when
+        // the bond moves to /v1/responses (verified 2026-08-28).
         codeExecutionToolType: 'code_interpreter',
         // LIST price. OpenAI ran a >20% PROMO from 2026-08-22 ($4/$20, cache read
         // $0.40, cache write $5) — "GPT-5.6 Sol's promotional pricing is available
@@ -479,7 +531,11 @@ export const MODELS = [
         supportsTools: true,
         // Tools + ANY reasoning is a 400 on /v1/chat/completions for this family.
         toolsRequireReasoningOff: true,
-        webSearchToolType: 'web_search',
+        // NO webSearchToolType: the OpenAI bond calls /v1/chat/completions, which
+        // has no web_search tool type (it is a Responses-API construct), and the
+        // bond deliberately forwards no server tools. Advertising one here surfaced
+        // web search in the system prompt while it could never work. Re-add when
+        // the bond moves to /v1/responses (verified 2026-08-28).
         codeExecutionToolType: 'code_interpreter',
         // Repriced 2026-07-30 (20% cut from $2.50/$15).
         inputPricePerMTok: 2,
@@ -507,7 +563,11 @@ export const MODELS = [
         supportsTools: true,
         // Tools + ANY reasoning is a 400 on /v1/chat/completions for this family.
         toolsRequireReasoningOff: true,
-        webSearchToolType: 'web_search',
+        // NO webSearchToolType: the OpenAI bond calls /v1/chat/completions, which
+        // has no web_search tool type (it is a Responses-API construct), and the
+        // bond deliberately forwards no server tools. Advertising one here surfaced
+        // web search in the system prompt while it could never work. Re-add when
+        // the bond moves to /v1/responses (verified 2026-08-28).
         codeExecutionToolType: 'code_interpreter',
         // Repriced 2026-07-30 (80% cut from $1/$6).
         inputPricePerMTok: 0.2,
@@ -542,7 +602,11 @@ export const MODELS = [
         supportsTools: true,
         // Tools + ANY reasoning is a 400 on /v1/chat/completions for this family.
         toolsRequireReasoningOff: true,
-        webSearchToolType: 'web_search',
+        // NO webSearchToolType: the OpenAI bond calls /v1/chat/completions, which
+        // has no web_search tool type (it is a Responses-API construct), and the
+        // bond deliberately forwards no server tools. Advertising one here surfaced
+        // web search in the system prompt while it could never work. Re-add when
+        // the bond moves to /v1/responses (verified 2026-08-28).
         codeExecutionToolType: 'code_interpreter',
         inputPricePerMTok: 5,
         outputPricePerMTok: 30,
@@ -572,7 +636,11 @@ export const MODELS = [
         supportsTools: true,
         // Tools + ANY reasoning is a 400 on /v1/chat/completions for this family.
         toolsRequireReasoningOff: true,
-        webSearchToolType: 'web_search',
+        // NO webSearchToolType: the OpenAI bond calls /v1/chat/completions, which
+        // has no web_search tool type (it is a Responses-API construct), and the
+        // bond deliberately forwards no server tools. Advertising one here surfaced
+        // web search in the system prompt while it could never work. Re-add when
+        // the bond moves to /v1/responses (verified 2026-08-28).
         codeExecutionToolType: 'code_interpreter',
         inputPricePerMTok: 2.5,
         outputPricePerMTok: 15,
@@ -604,7 +672,11 @@ export const MODELS = [
         supportsTools: true,
         // Tools + ANY reasoning is a 400 on /v1/chat/completions for this family.
         toolsRequireReasoningOff: true,
-        webSearchToolType: 'web_search',
+        // NO webSearchToolType: the OpenAI bond calls /v1/chat/completions, which
+        // has no web_search tool type (it is a Responses-API construct), and the
+        // bond deliberately forwards no server tools. Advertising one here surfaced
+        // web search in the system prompt while it could never work. Re-add when
+        // the bond moves to /v1/responses (verified 2026-08-28).
         codeExecutionToolType: 'code_interpreter',
         inputPricePerMTok: 0.75,
         outputPricePerMTok: 4.5,
@@ -1407,16 +1479,24 @@ export const MODELS = [
         supportedEffortLevels: ['4K', '8K', '16K', '32K'],
         defaultEffortLevel: '8K',
         effortBudgetTokens: { '4K': 4000, '8K': 8000, '16K': 16000, '32K': 32000 },
-        // models.dev claims image+video input, but Alibaba's own model catalog
-        // lists qwen3.8-max under text generation (VL remains a separate line) —
-        // false until the provider's page says otherwise.
+        // The provider's own model page DOES list Image/Text/Video input (checked
+        // 2026-08-28, superseding the earlier "text generation only" reading of the
+        // model catalog page). Kept false anyway: this model's DEFAULT region is the
+        // DeepInfra US re-host, which serves it as `text-generation` — vision is not
+        // offerable where it actually dispatches. Native-only qwen3.8-flash below
+        // carries the flag.
         supportsVision: false,
         supportsPromptCaching: true,
         supportsTools: true,
         inputPricePerMTok: 2,
         outputPricePerMTok: 6,
-        // Implicit context cache: read = 20% of input, no write premium.
-        cacheReadPricePerMTok: 0.4,
+        // Implicit context cache. The ZH cache doc's standard 20% does NOT apply
+        // here: that doc names qwen3.8-max as an exception and points at the
+        // console, and the model page publishes the real Singapore rate — $0.25,
+        // i.e. 12.5% of input (verified 2026-08-28 on
+        // alibabacloud.com/help/en/model-studio/qwen3-8-max). Implicit creation is
+        // not billed beyond input, so write stays 1× input.
+        cacheReadPricePerMTok: 0.25,
         cacheWritePricePerMTok: 2,
         regions: ['us', 'cn'],
         // US = DeepInfra (Qwen/Qwen3.8-Max), verified 2026-08-14 against
@@ -1430,6 +1510,49 @@ export const MODELS = [
         // Not published by Alibaba — best-effort estimate.
         knowledgeCutoff: '2026-04-01',
     },
+    {
+        id: 'qwen3.8-flash',
+        provider: 'alibaba',
+        label: 'Qwen3.8 Flash',
+        description: 'Alibaba cheap tier — 1M context, multimodal, hybrid thinking',
+        // Model page: context window 1,000,000; max input 991,808 (983,616 in
+        // thinking mode); max chain-of-thought 262,144; max output 131,072.
+        contextWindow: 1_000_000,
+        maxOutputTokens: 131_072,
+        // Same hybrid-thinking mechanism as qwen3.8-max — the deep-thinking doc
+        // lists the "Qwen3.8 Flash series" as hybrid with thinking ON by default,
+        // and thinking_budget applies — so effort scales the budget identically.
+        supportsThinking: true,
+        thinkingBudgetTokens: 8_000,
+        thinkingConfigurable: true,
+        supportedEffortLevels: ['4K', '8K', '16K', '32K'],
+        defaultEffortLevel: '8K',
+        effortBudgetTokens: { '4K': 4000, '8K': 8000, '16K': 16000, '32K': 32000 },
+        // Model page lists Image / Text / Video input. Unlike qwen3.8-max this one
+        // has no US re-host to lose it to — it only ever dispatches to the native
+        // host, which serves the multimodal surface.
+        supportsVision: true,
+        supportsPromptCaching: true,
+        supportsTools: true,
+        // Singapore International list card, published in USD on the model page:
+        // $0.15 / $0.47, implicit cache $0.016. (Beijing is cheaper at
+        // $0.113/$0.382/$0.014; the bond calls the international endpoint.)
+        inputPricePerMTok: 0.15,
+        outputPricePerMTok: 0.47,
+        // Implicit context cache. Like qwen3.8-max this model is an explicit
+        // exception to the ZH doc's standard 20%; the published rate is $0.016
+        // (~10.7% of input). Creation is not billed beyond input → write = 1×.
+        cacheReadPricePerMTok: 0.016,
+        cacheWritePricePerMTok: 0.15,
+        // No US re-host exists — DeepInfra serves Qwen3.8-Max, Qwen3.8-27B and
+        // Qwen3.8-2.4T-A95B but returns "model not found" for Qwen/Qwen3.8-Flash
+        // (checked 2026-08-28), so this is pinned to the native host and bills the
+        // card above. ALIBABA_US_MODEL_MAP therefore needs no entry.
+        regions: ['cn'],
+        // Not published by Alibaba — best-effort estimate, same 3.8 generation as
+        // qwen3.8-max.
+        knowledgeCutoff: '2026-04-01',
+    },
     {
         id: 'qwen3.7-max',
         provider: 'alibaba',
@@ -1583,6 +1706,11 @@ export const MODELS = [
         // US default: DeepInfra (zai-org/GLM-5.3-Flash) bills exactly the list
         // card — $0.15/$0.50, cache read 0.2× = $0.03. Verified live 2026-08-27.
         regions: ['us', 'cn'],
+        // NB: DeepInfra returns NO cached-token usage (verified live 2026-08-28:
+        // an identical ~2.6k-token prefix twice reported full input both passes,
+        // no prompt_tokens_details) — the us cacheRead rate below is informational
+        // only; metering always bills full input there. Native z.ai reports and
+        // discounts cached tokens correctly.
         regionPricing: {
             us: { inputPricePerMTok: 0.15, outputPricePerMTok: 0.5, cacheReadPricePerMTok: 0.03 },
         },
@@ -1614,6 +1742,11 @@ export const MODELS = [
         cacheWritePricePerMTok: 1.4,
         // US default (DeepInfra bills ~half native). Verified 2026-08-01.
         regions: ['us', 'cn'],
+        // NB: DeepInfra returns NO cached-token usage (verified live 2026-08-28:
+        // an identical ~2.6k-token prefix twice reported full input both passes,
+        // no prompt_tokens_details) — the us cacheRead rate below is informational
+        // only; metering always bills full input there. Native z.ai reports and
+        // discounts cached tokens correctly.
         regionPricing: {
             us: { inputPricePerMTok: 0.75, outputPricePerMTok: 2.4, cacheReadPricePerMTok: 0.14 },
         },
@@ -1650,6 +1783,11 @@ export const MODELS = [
         cacheWritePricePerMTok: 1,
         // US default (DeepInfra bills below native). Verified 2026-08-01.
         regions: ['us', 'cn'],
+        // NB: DeepInfra returns NO cached-token usage (verified live 2026-08-28:
+        // an identical ~2.6k-token prefix twice reported full input both passes,
+        // no prompt_tokens_details) — the us cacheRead rate below is informational
+        // only; metering always bills full input there. Native z.ai reports and
+        // discounts cached tokens correctly.
         regionPricing: {
             us: { inputPricePerMTok: 0.6, outputPricePerMTok: 2.08, cacheReadPricePerMTok: 0.12 },
         },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@molecule/api-resource-ai-models",
-  "version": "1.2.6",
+  "version": "1.3.0",
   "description": "AI model catalog — server-side source of truth plus an authentication-gated discovery endpoint",
   "type": "module",
   "main": "dist/index.js",