Remove four models from profit estimators / 从利润估算中移除四个模型

docs/tco-calculator.md

@@ -460,31 +460,11 @@ gigawatt-year buys for that SKU (`gw-year`).
 Both tabs offer the same model list. A model needs AgentX (`agentic_traces`)
 measurements with an interpolable operating point and a token price source.
 The page intersects its supported routes with live AgentX availability, so a
-model without current trace data disappears from the selector. Muse Spark Open,
-Qwen3.5, Qwen3.8 Flash Next, and MiniMax M3.1 Flash use OpenRouter pricing by
-default; users can supply custom prices. GLM-5 currently has too little trace
-data to interpolate a profit row.
+model without current trace data disappears from the selector.
 The public profit view APIs accept any known model and may return empty rows;
 their `model` parameter is not restricted to the dashboard's list.
 
-The four additions use model-specific default P90 interactivity targets. On
-2026-10-24, a sweep of the public profit view API at every 5 tok/s/user from 5
-to 150 found these lowest targets that maximize priced official SKU rows in
-that range (using custom token prices to keep price availability independent of
-curve coverage):
-
-| Model              | Rows at 45 | Default target (tok/s/user) | Rows at default |
-| ------------------ | ---------: | --------------------------: | --------------: |
-| Muse Spark Open    |          1 |                          85 |               7 |
-| MiniMax M3.1 Flash |          2 |                         125 |               7 |
-| Qwen3.5            |          2 |                         115 |              10 |
-| Qwen3.8 Flash Next |          3 |                          90 |               6 |
-
-Both per-chip-hour and per-GW-year views returned the same row counts at these
-targets. New benchmark runs can change the coverage. The target remains an
-editable operating-point assumption, not a hosted API speed claim.
-
-The five earlier models now follow the same rule. A 2026-10-25 public API sweep
+Each model uses a default P90 interactivity target. A 2026-10-25 public API sweep
 at every 5 tok/s/user from 5 to 300 found these lowest targets that maximize
 priced official SKU rows in that range, again using custom token prices to
 isolate curve coverage:

packages/app/cypress/e2e/profit-estimator.cy.ts

@@ -13,10 +13,6 @@ import {
 
 const OPENROUTER_MODELS = {
   data: [
-    {
-      id: 'meta/muse-spark-open',
-      pricing: { prompt: '0.00000125', completion: '0.00000425', input_cache_read: '0.00000015' },
-    },
     {
       id: 'moonshotai/kimi-k3',
       pricing: { prompt: '0.0000006', completion: '0.0000025', input_cache_read: '0.0000001' },
@@ -37,18 +33,6 @@ const OPENROUTER_MODELS = {
       id: 'deepseek/deepseek-v4.1-flash',
       pricing: { prompt: '0.00000015', completion: '0.0000006', input_cache_read: '0.000000003' },
     },
-    {
-      id: 'minimax/minimax-m3.1-flash',
-      pricing: { prompt: '0.00000015', completion: '0.0000006', input_cache_read: '0.00000003' },
-    },
-    {
-      id: 'qwen/qwen3.5-397b-a17b',
-      pricing: { prompt: '0.0000005', completion: '0.0000036', input_cache_read: '0.00000005' },
-    },
-    {
-      id: 'qwen/qwen3.8-flash',
-      pricing: { prompt: '0.00000015', completion: '0.00000047', input_cache_read: '0.00000002' },
-    },
   ],
 };
 
@@ -85,31 +69,6 @@ function coverageRows(dbKey: string, target: number) {
   });
 }
 
-describe('New agentic models on profit estimators', () => {
-  for (const [slug, label, dbKey, target] of [
-    ['muse-spark-open', 'Muse Spark Open', 'musespark', 85],
-    ['minimax-m31-flash', 'MiniMax M3.1 Flash', 'minimaxm31flash', 125],
-    ['qwen-3-5', 'Qwen3.5', 'qwen3.5', 115],
-    ['qwen-3-8-flash-next', 'Qwen3.8 Flash Next', 'qwen3.8next', 90],
-  ] as const) {
-    for (const basis of ['profit-estimator', 'profit-estimator-per-gigawatt'] as const) {
-      it(`opens ${label} on ${basis} with more measured SKUs than target 45`, () => {
-        stubOpenRouter();
-        cy.intercept('GET', '/api/v1/benchmarks*', { body: coverageRows(dbKey, target) });
-        cy.visit(`/${basis}/${slug}`, { onBeforeLoad: suppressNudges });
-        cy.location('pathname').should('eq', `/${basis}/${slug}`);
-        cy.get('[data-testid="profit-model-selector"]').should('contain.text', label);
-        cy.get('button#profit-price-source').should('contain.text', 'OpenRouter');
-        cy.get('[data-testid="profit-target-input"]').should('have.value', String(target));
-        chart().find('text.revenue-label').should('have.length', 4);
-
-        cy.get('[data-testid="profit-target-input"]').clear().type('45').blur();
-        chart().find('text.revenue-label').should('have.length', 1);
-      });
-    }
-  }
-});
-
 describe('Earlier agentic models on profit estimators', () => {
   for (const [slug, label, dbKey, previousTarget, target, previousBars] of [
     ['kimi-k3', 'Kimi K3', 'kimik3', 45, 60, 1],
@@ -639,16 +598,12 @@ describe('Profit Estimator per GW', () => {
     cy.get('[data-testid="profit-precision-selector"]').should('not.exist');
     cy.get('[data-testid="profit-model-selector"]').should('contain.text', 'Kimi K3').click();
     cy.get('[role="option"]')
-      .should('have.length', 9)
-      .and('contain.text', 'Muse Spark Open')
+      .should('have.length', 5)
       .and('contain.text', 'Kimi K3')
       .and('contain.text', 'GLM5.2/GLM5.3')
       .and('contain.text', 'MiniMax M3')
       .and('contain.text', 'DeepSeek V4 Pro')
-      .and('contain.text', 'DeepSeek V4.1 Flash')
-      .and('contain.text', 'MiniMax M3.1 Flash')
-      .and('contain.text', 'Qwen3.5')
-      .and('contain.text', 'Qwen3.8 Flash Next');
+      .and('contain.text', 'DeepSeek V4.1 Flash');
     cy.get('body').type('{esc}');
     cy.get('button#profit-price-source').click();
     cy.get('[role="option"]')
@@ -1148,9 +1103,12 @@ describe('Profit Estimator per GW — per-model route', () => {
         cy.get('[data-testid="profit-scenario-selector"]').should('not.exist');
         cy.get('[data-testid="profit-fixed-workload-note"]').should('not.exist');
       });
-      cy.request(`${prefix}/profit-estimator-per-gigawatt/qwen-3-5`)
+      cy.request({
+        url: `${prefix}/profit-estimator-per-gigawatt/qwen-3-5`,
+        failOnStatusCode: false,
+      })
         .its('status')
-        .should('eq', 200);
+        .should('eq', 404);
     });
   }
 });

packages/app/cypress/support/profit-fixtures.ts

@@ -6,15 +6,11 @@ export const PROFIT_MINIMAX_DB_KEY = 'minimaxm3';
 export const PROFIT_DEEPSEEK_DB_KEY = 'dsv4';
 export const PROFIT_DEEPSEEK_FLASH_DB_KEY = 'dsv41flash';
 const PROFIT_DB_KEY_BY_DISPLAY_MODEL: Record<string, string> = {
-  'Muse-Spark-Open': 'musespark',
   'Kimi-K3': PROFIT_MODEL_DB_KEY,
   'GLM-5.2': PROFIT_GLM_DB_KEY,
   'MiniMax-M3': PROFIT_MINIMAX_DB_KEY,
   'DeepSeek-V4-Pro': PROFIT_DEEPSEEK_DB_KEY,
   'DeepSeek-V4.1-Flash': PROFIT_DEEPSEEK_FLASH_DB_KEY,
-  'MiniMax-M3.1-Flash': 'minimaxm31flash',
-  'Qwen-3.5-397B-A17B': 'qwen3.5',
-  'Qwen3.8-Flash-Next': 'qwen3.8next',
 };
 const PROFIT_DB_KEYS = Object.values(PROFIT_DB_KEY_BY_DISPLAY_MODEL);
 export const PROFIT_DATE = '2026-08-31';

packages/app/src/app/api/v1/views/extensions.test.ts

@@ -327,10 +327,6 @@ describe('new dashboard projections', () => {
     [Model.MiniMax_M3, 'minimaxm3', 115],
     [Model.DeepSeek_V4_Pro, 'dsv4', 170],
     [Model.DeepSeek_V4_1_Flash, 'dsv41flash', 150],
-    [Model.Muse_Spark_Open, 'musespark', 85],
-    [Model.MiniMax_M3_1_Flash, 'minimaxm31flash', 125],
-    [Model.Qwen3_5, 'qwen3.5', 115],
-    [Model.Qwen3_8_Flash_Next, 'qwen3.8next', 90],
   ] as const)(
     'prices supported agentic model %s at its default through both public profit views',
     async (model, dbKey, target) => {

packages/app/src/components/calculator/profit-estimator.test.ts

@@ -333,19 +333,6 @@ describe('parseTokenPriceInput', () => {
 });
 
 describe('profitModelDefaults', () => {
-  it.each([
-    [Model.Muse_Spark_Open, 85],
-    [Model.MiniMax_M3_1_Flash, 125],
-    [Model.Qwen3_5, 115],
-    [Model.Qwen3_8_Flash_Next, 90],
-  ] as const)('opens %s at a 5-step target chosen for broader SKU coverage', (model, target) => {
-    const defaults = profitModelDefaults(model);
-    expect(defaults.interactivity).toBe(target);
-    expect(defaults.interactivity % 5).toBe(0);
-    expect(defaults.listPricing).toBeNull();
-    expect(defaults.labCutPct).toBe(DEFAULT_LAB_CUT_PCT);
-  });
-
   it.each([
     [Model.Kimi_K3, 60],
     [Model.GLM_5_2, 135],

packages/app/src/components/calculator/profit-estimator.ts

@@ -61,10 +61,8 @@ export interface ProfitModelDefaults {
 }
 
 /**
- * Per-model defaults. On 2026-10-24, the four newest models took the first
- * 5-tok/s/user step that maximized priced AgentX SKU coverage in a 5–150 sweep:
- * Muse Spark Open 85, MiniMax M3.1 Flash 125, Qwen3.5 115, Qwen3.8 Flash Next 90.
- * On 2026-10-25, the same rule over 5–300 selected Kimi K3 60, GLM 5.2/5.3
+ * Per-model defaults. On 2026-10-25, the first 5-tok/s/user step that
+ * maximized priced AgentX SKU coverage in a 5–300 sweep selected Kimi K3 60, GLM 5.2/5.3
  * 135, MiniMax M3 115, DeepSeek V4 Pro 170, and DeepSeek V4.1 Flash 150.
  * These editable P90 targets show measured configurations; they are not claims
  * about hosted API speeds. Curves outside the target range stay unpriced.
@@ -77,26 +75,6 @@ export interface ProfitModelDefaults {
  * list prices and a 0% fee because their weights ship under the MIT license.
  */
 const PROFIT_MODEL_DEFAULTS: Partial<Record<Model, ProfitModelDefaults>> = {
-  [Model.Muse_Spark_Open]: {
-    interactivity: 85,
-    labCutPct: DEFAULT_LAB_CUT_PCT,
-    listPricing: null,
-  },
-  [Model.MiniMax_M3_1_Flash]: {
-    interactivity: 125,
-    labCutPct: DEFAULT_LAB_CUT_PCT,
-    listPricing: null,
-  },
-  [Model.Qwen3_5]: {
-    interactivity: 115,
-    labCutPct: DEFAULT_LAB_CUT_PCT,
-    listPricing: null,
-  },
-  [Model.Qwen3_8_Flash_Next]: {
-    interactivity: 90,
-    labCutPct: DEFAULT_LAB_CUT_PCT,
-    listPricing: null,
-  },
   [Model.DeepSeek_V4_Pro]: {
     interactivity: 170,
     labCutPct: 0,

packages/app/src/lib/api/api-route-catalog.ts

@@ -1361,7 +1361,7 @@ export const apiContractSourceDigests = [
   },
   {
     source: 'src/components/calculator/profit-estimator.ts',
-    sourceSha256: '7ccc4a040faef875fef299bd4c29f33645a914efc69f9b73b53cda676d7b4aa7',
+    sourceSha256: '0eeaaed822068e9aa782e6b40ad1bb9ae31f70ef755f67622767f3b89ade4d1a',
     reviewArea: {
       en: 'Shared profit-estimator defaults and calculation parity, including model-specific target assumptions.',
       zh: '利润估算器的共享默认值与计算一致性，包括各模型的目标假设。',

packages/app/src/lib/catalog/data-mappings.test.ts

@@ -396,12 +396,15 @@ describe('getModelLabel', () => {
 });
 
 describe('getOpenRouterModelId', () => {
-  it('maps catalog models to exact OpenRouter ids', () => {
+  it('maps catalog models to exact OpenRouter ids without inventing one for Muse', () => {
     for (const model of MODEL_OPTIONS) {
+      if (model === Model.Muse_Spark_Open) {
+        expect(getOpenRouterModelId(model)).toBeNull();
+        continue;
+      }
       expect(getOpenRouterModelId(model), model).toMatch(/^[^/]+\/[^/]+$/u);
     }
     expect(getOpenRouterModelId(Model.DeepSeek_V4_Pro)).toBe('deepseek/deepseek-v4-pro-0813');
-    expect(getOpenRouterModelId(Model.Muse_Spark_Open)).toBe('meta/muse-spark-open');
   });
 });
 

packages/app/src/lib/catalog/data-mappings.ts

@@ -74,7 +74,6 @@ const MODEL_CONFIG: Record<Model, ModelConfig> = {
     label: 'Muse Spark Open 816B',
     prefix: 'musespark',
     category: 'default',
-    openRouterModelId: 'meta/muse-spark-open',
     logo: 'muse-mascot.webp',
   },
   [Model.DeepSeek_V4_Pro]: {

packages/app/src/lib/routing/model-routes.test.ts

@@ -175,23 +175,16 @@ describe('modelRoutePathnameRewrite', () => {
 });
 
 describe('modelRoutesForTab', () => {
-  it('serves agentic models with priced interpolation on both profit estimators and every model elsewhere', () => {
+  it('serves Kimi K3, GLM 5.2/5.3, MiniMax M3, DeepSeek V4 Pro and DeepSeek V4.1 Flash on the profit estimators and every model elsewhere', () => {
     for (const tab of ['profit-estimator', 'profit-estimator-per-gigawatt'] as const) {
       expect(modelRoutesForTab(tab).map((route) => route.model)).toEqual([
-        Model.Muse_Spark_Open,
         Model.DeepSeek_V4_Pro,
         Model.DeepSeek_V4_1_Flash,
         Model.Kimi_K3,
         Model.MiniMax_M3,
-        Model.MiniMax_M3_1_Flash,
         Model.GLM_5_2,
-        Model.Qwen3_5,
-        Model.Qwen3_8_Flash_Next,
       ]);
-      expect(modelRouteAvailableForTab(tab, Model.Qwen3_5)).toBe(true);
-      expect(modelRouteAvailableForTab(tab, Model.Qwen3_8_Flash_Next)).toBe(true);
-      expect(modelRouteAvailableForTab(tab, Model.MiniMax_M3_1_Flash)).toBe(true);
-      expect(modelRouteAvailableForTab(tab, Model.Muse_Spark_Open)).toBe(true);
+      expect(modelRouteAvailableForTab(tab, Model.Qwen3_5)).toBe(false);
       expect(modelRouteAvailableForTab(tab, Model.GLM_5_2)).toBe(true);
       expect(modelRouteAvailableForTab(tab, Model.MiniMax_M3)).toBe(true);
       expect(modelRoutesForTab(tab).find((route) => route.model === Model.MiniMax_M3)?.slug).toBe(

packages/app/src/lib/routing/model-routes.ts

@@ -34,15 +34,11 @@ export function defaultRouteModel(tab: ModelRouteTab): Model {
 }
 
 const PROFIT_ESTIMATOR_MODELS: readonly Model[] = [
-  Model.Muse_Spark_Open,
   Model.Kimi_K3,
   Model.GLM_5_2,
   Model.MiniMax_M3,
-  Model.MiniMax_M3_1_Flash,
   Model.DeepSeek_V4_Pro,
   Model.DeepSeek_V4_1_Flash,
-  Model.Qwen3_5,
-  Model.Qwen3_8_Flash_Next,
 ];
 const MODEL_ROUTE_TAB_MODELS: Partial<Record<ModelRouteTab, readonly Model[]>> = {
   'profit-estimator': PROFIT_ESTIMATOR_MODELS,

packages/app/src/lib/views-api/docs/extensions.ts

@@ -72,8 +72,8 @@ const PARAMETER_NOTES: Record<string, [string, string]> = {
     '比较范围结束日期 YYYY-MM-DD，不得早于 start。历史视图将其用作含终点的数据范围。',
   ],
   target: [
-    'Positive operating-point target, tok/s/user in interactivity-to-throughput mode. Both profit views use model-specific defaults: Kimi K3 60, GLM-5.2/5.3 135, MiniMax M3 115, DeepSeek V4 Pro 170, DeepSeek V4.1 Flash 150, Muse Spark Open 85, MiniMax M3.1 Flash 125, Qwen3.5 115, and Qwen3.8 Flash Next 90 tok/s/user. Dashboard links use /<model>/target/<value> route segments; this API continues to accept the target query parameter. These are editable SKU-coverage assumptions, not OpenRouter P90-equivalent measurements.',
-    '正数工作点目标；交互性转吞吐量模式下单位为 tok/s/user。两个利润视图按模型设置默认值：Kimi K3 为 60、GLM-5.2/5.3 为 135、MiniMax M3 为 115、DeepSeek V4 Pro 为 170、DeepSeek V4.1 Flash 为 150、Muse Spark Open 为 85、MiniMax M3.1 Flash 为 125、Qwen3.5 为 115、Qwen3.8 Flash Next 为 90 tok/s/user。仪表板链接通过 /<model>/target/<value> 路径片段传递目标值；本 API 仍使用 target 查询参数。这些默认值是用于扩大 SKU 覆盖范围的可修改假设，并非与 OpenRouter P90 等价的实测值。',
+    'Positive operating-point target, tok/s/user in interactivity-to-throughput mode. Both profit views use model-specific defaults: Kimi K3 60, GLM-5.2/5.3 135, MiniMax M3 115, DeepSeek V4 Pro 170 and DeepSeek V4.1 Flash 150 tok/s/user. Dashboard links use /<model>/target/<value> route segments; this API continues to accept the target query parameter. These are editable SKU-coverage assumptions, not OpenRouter P90-equivalent measurements.',
+    '正数工作点目标；交互性转吞吐量模式下单位为 tok/s/user。两个利润视图按模型设置默认值：Kimi K3 为 60、GLM-5.2/5.3 为 135、MiniMax M3 为 115、DeepSeek V4 Pro 为 170、DeepSeek V4.1 Flash 为 150 tok/s/user。仪表板链接通过 /<model>/target/<value> 路径片段传递目标值；本 API 仍使用 target 查询参数。这些默认值是用于扩大 SKU 覆盖范围的可修改假设，并非与 OpenRouter P90 等价的实测值。',
   ],
   costProvider: [
     'costh = owning, costr = renting. Profit views additionally accept custom with customCosts. No costn provider.',

packages/skills/skills/inferencex-api/integrity.json

@@ -9,7 +9,7 @@
     "references/cli-contract.md": "4f070877eeefffad2860c58f79f68432fbe27aa21fc578a1c696f0309a3d7d7b",
     "references/cli.md": "11e1183f7ed70d669019e8435ac16a6a9f873de027507b8850a9943f6659bfd9",
     "references/collectivex.md": "f79cf8e2779a9289ab69f79f4a4b338e0b9a5f76a3750a3e969f33a744e726a1",
-    "references/dashboard-views.md": "333e4473a6027a45992a01bc2e908437dc6b5c43542dd481b0929ba95027f506",
+    "references/dashboard-views.md": "db9ba48896f75acf6531bd49d26062ee09307c6348f1e10b055cf62e093df85d",
     "references/offline-exports.md": "95aa565dcd4c9e592159baa560a9a58217bc61e395de1c6daf0576795c7f4ec9",
     "references/pareto.md": "1b4d2d163f982e3f2d97addce310789eae9c1e5badd82dd7501708c9e0385027",
     "references/powerx.md": "5af5527d54a7e6191a5769080c122638b333b1ceb59e19fdee7df662a5cd3c8f",