slash-tokens 1.6.4 → 1.6.6

package/CHANGELOG.md

@@ -1,10 +1,43 @@
 # Changelog
 
+## [1.6.6] — The Fixed Deal Edition
+
+*2026-10-07*
+
+`--version` / `--help` answer, and today's models price correctly.
+
+Solo $20 mailbox, 10% waived. Team $39 for the data.
+
+### Fixed
+- **`slash-tokens --version` and `--help` print and exit.** In 1.6.5 every flag fell through to a full scan of the current folder, so `--version` run from `~` scanned the home folder. An empty `bunx` run now tells you to run it in an app.
+- **`preflight()` no longer throws on current models or real API IDs.** IDs are canonicalised strictly (`claude-opus-4-7` → `claude-opus-4.7`, date stamps dropped). There is no family guessing: an unknown version still fails with "Unknown model", so a new model never silently gets an older model's price.
+- **`npm publish` always builds first** (`prepublishOnly`), so a stale `dist/` can't ship again (1.6.4).
+- **`report()` sends numbers only.** A test pins its payload to `tokens_estimated`, `tokens_saved`, `model`, `action`, `cost_saved_usd`; prompt content never leaves the machine.
+- **`npm test` runs offline.** The live integration suite (`tests/z-integration.test.ts`, which registers keys on mcpaas.live) runs only with `SLASH_LIVE=1`; the weekly `integration.yml` sets it.
+
+### Changed
+- **Model table, prices as of 2026-10-07** (checked against the Anthropic, OpenAI, xAI and Google pricing pages). Added Claude Opus 5.5 ($4/$20), Sonnet 5.5 ($2/$10), Opus 4.8 / 4.6 / 4.5, Sonnet 4.6 / 4.5, Fable 5 / 5.1 and Mythos 5 / 5.1 ($10/$50); Grok 4.7 and 4.5 ($2/$6, doubling above 200K), grok-build-0.1 ($1/$2); GPT-6 Astra ($10/$50), GPT-6.1 Sol and GPT-6 Sol ($2/$10), GPT-6 Luna ($0.10/$0.50); Gemini 3.8 / 3.7 / 3.6 Flash ($0.75/$3.75 launch price through 2026-12-31), 3.5 Flash ($1.50/$9), 3.1 Flash-Lite ($0.25/$1.50), 3.1 Pro Preview.
+- **`slash-tokens/auto` routing is unchanged within 1.6.x.** `/auto` rewrites live requests only to the models it routed to in 1.6.5 (`AUTO_ROUTE_TARGETS`), using the same identification, so a patch upgrade never changes where production calls go. New models are now recognised and priced correctly in `/auto` events (`identifyModel`: `claude-opus-5-5` at $4/$20, `gpt-6-astra` at $10/$50); /auto starts routing to them in 1.7.0.
+- **Routing (`preflightRoute`) follows the new ladder.** Cheapest same-provider is now GPT-6 Luna for OpenAI and Gemini 3.1 Flash-Lite for Google (Anthropic and xAI unchanged). grok-build-0.1 is priced but never a routing target (`NOT_ROUTE_TARGETS`): it's a coding-agent model. When two alternatives' costs round to the same value (tiny prompts), the lower list price wins.
+- **Calibration:** new Claude, Grok and Gemini models use their family's measured factor. GPT-6 is a new generation with an unbenchmarked tokenizer, so it takes the conservative default until a bench run adds it (slash never under-reports).
+
+No Team/Solo price change.
+
+## [1.6.5] — The Fixed Deal Edition
+
+*2026-08-25*
+
+Solo $20 mailbox, 10% waived. Team $39 for the data.
+
+Rebuilt tarball. 1.6.4 packed a stale gitignored `dist/` (`bun test` runs `src/`). This tarball contains the live ladder: Grok **4.6** / **4.3**, GPT-5.6 **Sol / Terra / Luna**, Claude **Opus 5 / Sonnet 5 / Haiku 4.5**, Gemini **3.5 Flash-Lite**.
+
+No Team/Solo price change.
+
 ## [1.6.4] — The Fixed Deal Edition
 
 *2026-08-25*
 
-Live model ladder. Grok **4.6** / **4.3**, GPT-5.6 **Sol / Terra / Luna**, Claude **Opus 5 / Sonnet 5 / Haiku 4.5**, Gemini **3.5 Flash-Lite**. Old keys stay as aliases. Calibration factors carried from the 2026-08-23 corpus — not re-measured on the new IDs.
+Live model ladder in source. **Tarball packed stale `dist/`** — `bun test` never rebuilds it. Use **1.6.5**.
 
 No Team/Solo price change.
 

package/README.md

@@ -15,10 +15,14 @@ Know the cost before the call leaves your machine.
 Models change. Windows grow. Slash adapts — you keep building.
 Cheaper tokens haven't shrunk the bill — usage has.
 
-## v1.6.4 — The Fixed Deal Edition
+## v1.6.6 — The Fixed Deal Edition
+
+`--version` / `--help` answer, and today's models price correctly.
 
 Solo $20 mailbox, 10% waived. Team $39 for the data.
 
+New in 1.6.6: Claude Opus 5.5 / Sonnet 5.5 / Fable 5.1, Grok 4.7, GPT-6 (Astra, Sol, Luna), Gemini 3.6–3.8 Flash and 3.1 Flash-Lite, priced as of 2026-10-07. Real API IDs (`claude-opus-4-7`) work in `preflight()`. `preflightRoute()` now finds GPT-6 Luna and Gemini 3.1 Flash-Lite as the cheapest same-provider options.
+
 **Free forever is bunx** — no account. A one-person account is email → key, **$20 on the house**. We show the savings. We don't charge. 10% is the model, waived. Team is **$39 for the data** (`$390`/year).
 
 ```bash
@@ -26,7 +30,7 @@ bunx slash-tokens
 # or: npx --yes slash-tokens
 ```
 
-Run it in a project that already calls an LLM. An empty folder prints that nothing was found — that's normal.
+Run it in a project that already calls an LLM. An empty folder prints that nothing was found, then tells you to run it in an app — that's normal. `--version` / `--help` print and exit (they do not scan). Pin proof: `slash-tokens --version` or `npm view slash-tokens version`.
 
 ```bash
 npm install slash-tokens

package/dist/cli.js

@@ -1,5 +1,10 @@
 #!/usr/bin/env node
 
+// src/cli.ts
+import { readFileSync as readFileSync2 } from "node:fs";
+import { dirname, join as join2 } from "node:path";
+import { fileURLToPath } from "node:url";
+
 // src/scanner.ts
 import { readdirSync, readFileSync, statSync } from "fs";
 import { join, extname } from "path";
@@ -49,17 +54,162 @@ function writeToMemory(content) {
   return maxLen;
 }
 
+// src/models.ts
+var OPUS = { input: 5, output: 25, context: 1e6 };
+var OPUS_55 = { input: 4, output: 20, context: 1e6 };
+var FABLE = { input: 10, output: 50, context: 1e6 };
+var SONNET_4X = { input: 3, output: 15, context: 1e6 };
+var SONNET = { input: 2, output: 10, context: 1e6 };
+var HAIKU = { input: 1, output: 5, context: 200000 };
+var GROK_46 = {
+  input: 2,
+  output: 6,
+  context: 500000,
+  longContextThreshold: 200000,
+  longContextInput: 4,
+  longContextOutput: 12
+};
+var GROK_43 = {
+  input: 1.25,
+  output: 2.5,
+  context: 1e6,
+  longContextThreshold: 200000,
+  longContextInput: 2.5,
+  longContextOutput: 5
+};
+var GEMINI_PRO = {
+  input: 2,
+  output: 12,
+  context: 1e6,
+  longContextThreshold: 200000,
+  longContextInput: 4,
+  longContextOutput: 18
+};
+var GEMINI_FLASH = { input: 0.3, output: 2.5, context: 1e6 };
+var GEMINI_FLASH_3X = { input: 0.75, output: 3.75, context: 1e6 };
+var GEMINI_35_FLASH = { input: 1.5, output: 9, context: 1e6 };
+var GEMINI_31_FLASH_LITE = { input: 0.25, output: 1.5, context: 1e6 };
+var GROK_BUILD = {
+  input: 1,
+  output: 2,
+  context: 256000,
+  longContextThreshold: 200000,
+  longContextInput: 2,
+  longContextOutput: 4
+};
+var GPT_6_ASTRA = { input: 10, output: 50, context: 1050000 };
+var GPT_6_SOL = { input: 2, output: 10, context: 1050000 };
+var GPT_6_LUNA = { input: 0.1, output: 0.5, context: 1050000 };
+var GPT_SOL = { input: 4, output: 20, context: 1050000 };
+var GPT_TERRA = { input: 2, output: 12, context: 1050000 };
+var GPT_LUNA = { input: 0.2, output: 1.2, context: 1050000 };
+var GPT_54 = { input: 2.5, output: 15, context: 1e6 };
+var GPT_54_MINI = { input: 0.75, output: 4.5, context: 128000 };
+var GPT_54_NANO = { input: 0.2, output: 1.25, context: 128000 };
+var MODELS = {
+  "claude-fable-5.1": { ...FABLE },
+  "claude-fable-5": { ...FABLE },
+  "claude-mythos-5.1": { ...FABLE },
+  "claude-mythos-5": { ...FABLE },
+  "claude-opus-5.5": { ...OPUS_55 },
+  "claude-opus-5": { ...OPUS },
+  "claude-opus-4.8": { ...OPUS },
+  "claude-opus": { ...OPUS },
+  "claude-opus-4.7": { ...OPUS },
+  "claude-opus-4.6": { ...OPUS },
+  "claude-opus-4.5": { ...OPUS },
+  "claude-sonnet-5.5": { ...SONNET },
+  "claude-sonnet-5": { ...SONNET },
+  "claude-sonnet": { ...SONNET },
+  "claude-sonnet-4.6": { ...SONNET_4X },
+  "claude-sonnet-4.5": { ...SONNET_4X },
+  "claude-haiku-4.5": { ...HAIKU },
+  "claude-haiku": { ...HAIKU },
+  "grok-4.7": { ...GROK_46 },
+  "grok-4.6": { ...GROK_46 },
+  "grok-4.5": { ...GROK_46 },
+  "grok-build-0.1": { ...GROK_BUILD },
+  "grok-4.3": { ...GROK_43 },
+  "grok-4.20": { ...GROK_43 },
+  "grok-4-1-fast": { ...GROK_43 },
+  "gemini-3.1-pro": { ...GEMINI_PRO },
+  "gemini-3.1-pro-preview": { ...GEMINI_PRO },
+  "gemini-3.8-flash": { ...GEMINI_FLASH_3X },
+  "gemini-3.7-flash": { ...GEMINI_FLASH_3X },
+  "gemini-3.6-flash": { ...GEMINI_FLASH_3X },
+  "gemini-3.5-flash": { ...GEMINI_35_FLASH },
+  "gemini-3.1-flash-lite": { ...GEMINI_31_FLASH_LITE },
+  "gemini-3.5-flash-lite": { ...GEMINI_FLASH },
+  "gemini-2.5-flash": { ...GEMINI_FLASH },
+  "gpt-6-astra": { ...GPT_6_ASTRA },
+  "gpt-6.1-sol": { ...GPT_6_SOL },
+  "gpt-6-sol": { ...GPT_6_SOL },
+  "gpt-6-luna": { ...GPT_6_LUNA },
+  "gpt-5.6-sol": { ...GPT_SOL },
+  "gpt-5.6-terra": { ...GPT_TERRA },
+  "gpt-5.6-luna": { ...GPT_LUNA },
+  "gpt-5.4": { ...GPT_54 },
+  "gpt-5.4-mini": { ...GPT_54_MINI },
+  "gpt-5.4-nano": { ...GPT_54_NANO }
+};
+function canonicalModel(name) {
+  if (MODELS[name])
+    return name;
+  let n = name.trim().toLowerCase();
+  if (MODELS[n])
+    return n;
+  n = n.replace(/-\d{8}$/, "");
+  if (MODELS[n])
+    return n;
+  const dotted = n.replace(/(\d)-(\d)(?=$|-)/g, "$1.$2");
+  if (MODELS[dotted])
+    return dotted;
+  return n;
+}
+function getModel(name) {
+  return MODELS[canonicalModel(name)];
+}
+
 // src/slash.ts
 var WASM_INPUT_OFFSET2 = 4096;
 var CALIBRATION = {
+  "claude-fable-5.1": 2.05,
+  "claude-fable-5": 2.05,
+  "claude-mythos-5.1": 2.05,
+  "claude-mythos-5": 2.05,
+  "claude-opus-5.5": 2.05,
+  "claude-opus-5": 2.05,
+  "claude-opus-4.8": 2.05,
+  "claude-opus-4.6": 2.05,
+  "claude-opus-4.5": 2.05,
+  "claude-sonnet-5.5": 2.05,
+  "claude-sonnet-4.6": 2.05,
+  "claude-sonnet-4.5": 2.05,
   "claude-opus": 2.05,
   "claude-opus-4.7": 2.05,
+  "claude-sonnet-5": 2.05,
   "claude-sonnet": 2.05,
+  "claude-haiku-4.5": 1.45,
   "claude-haiku": 1.45,
   "gemini-3.1-pro": 1.45,
+  "gemini-3.1-pro-preview": 1.45,
+  "gemini-3.8-flash": 1.45,
+  "gemini-3.7-flash": 1.45,
+  "gemini-3.6-flash": 1.45,
+  "gemini-3.5-flash": 1.45,
+  "gemini-3.1-flash-lite": 1.45,
+  "gemini-3.5-flash-lite": 1.45,
   "gemini-2.5-flash": 1.45,
+  "grok-4.7": 1.15,
+  "grok-4.6": 1.15,
+  "grok-4.5": 1.15,
+  "grok-build-0.1": 1.15,
+  "grok-4.3": 1.15,
   "grok-4.20": 1.15,
   "grok-4-1-fast": 1.15,
+  "gpt-5.6-sol": 1.15,
+  "gpt-5.6-terra": 1.15,
+  "gpt-5.6-luna": 1.15,
   "gpt-5.4": 1.15,
   "gpt-5.4-mini": 1.15,
   "gpt-5.4-nano": 1.15
@@ -73,7 +223,7 @@ function slash(content, model) {
   const raw = instance.exports.estimate_tokens(WASM_INPUT_OFFSET2, len);
   if (!model)
     return raw;
-  const factor = CALIBRATION[model] ?? DEFAULT_UNKNOWN_MODEL_FACTOR;
+  const factor = CALIBRATION[canonicalModel(model)] ?? DEFAULT_UNKNOWN_MODEL_FACTOR;
   return factor === 1 ? raw : Math.ceil(raw * factor);
 }
 
@@ -91,12 +241,12 @@ var AI_PATTERNS = [
   { name: "Mistral", regex: /from\s+['"]@mistralai|MistralClient/g }
 ];
 var SDK_REPRESENTATIVE_MODEL = {
-  Anthropic: "claude-sonnet",
-  OpenAI: "gpt-5.4",
+  Anthropic: "claude-sonnet-5",
+  OpenAI: "gpt-5.6-sol",
   Gemini: "gemini-3.1-pro",
-  Grok: "grok-4.20"
+  Grok: "grok-4.6"
 };
-var UNKNOWN_SDK_REPRESENTATIVE_MODEL = "claude-sonnet";
+var UNKNOWN_SDK_REPRESENTATIVE_MODEL = "claude-sonnet-5";
 var SKIP_DIRS = new Set([
   "node_modules",
   ".git",
@@ -202,24 +352,6 @@ function scan(dir) {
   };
 }
 
-// src/models.ts
-var MODELS = {
-  "claude-opus": { input: 5, output: 25, context: 1e6 },
-  "claude-opus-4.7": { input: 5, output: 25, context: 1e6 },
-  "claude-sonnet": { input: 2, output: 10, context: 1e6 },
-  "claude-haiku": { input: 1, output: 5, context: 200000 },
-  "grok-4.20": { input: 1.25, output: 2.5, context: 1e6, longContextThreshold: 200000, longContextInput: 2.5, longContextOutput: 5 },
-  "grok-4-1-fast": { input: 1.25, output: 2.5, context: 1e6, longContextThreshold: 200000, longContextInput: 2.5, longContextOutput: 5 },
-  "gemini-3.1-pro": { input: 2, output: 12, context: 1e6 },
-  "gemini-2.5-flash": { input: 0.3, output: 2.5, context: 1e6 },
-  "gpt-5.4": { input: 2.5, output: 15, context: 1e6 },
-  "gpt-5.4-mini": { input: 0.75, output: 4.5, context: 128000 },
-  "gpt-5.4-nano": { input: 0.2, output: 1.25, context: 128000 }
-};
-function getModel(name) {
-  return MODELS[name] || MODELS[name.toLowerCase()];
-}
-
 // src/report.ts
 var R = "\x1B[0m";
 var B = "\x1B[1m";
@@ -244,6 +376,7 @@ function printReport(sites, filesScanned, timeMs, cwd) {
   if (sites.length === 0) {
     console.log(`${WHITE}  No AI API call sites detected.${R}`);
     console.log(`${GRAY}  Supported: OpenAI, Anthropic, Vercel AI, LangChain, Gemini, Bedrock, Grok${R}`);
+    console.log(`${GRAY}  Run this in a project that already calls an LLM.${R}`);
     console.log("");
     return;
   }
@@ -315,6 +448,14 @@ function init(opts) {
 
 // src/cli.ts
 var args = process.argv.slice(2);
+if (args.includes("--version") || args.includes("-V")) {
+  console.log(packageVersion());
+  process.exit(0);
+}
+if (args.includes("--help") || args.includes("-h")) {
+  printHelp();
+  process.exit(0);
+}
 var keyArg = args.find((a) => a.startsWith("--key="));
 var key = keyArg?.split("=")[1] || process.env.SLASH_KEY;
 if (key)
@@ -336,3 +477,21 @@ if (sites.length > 0) {
   }
   console.log("");
 }
+function packageVersion() {
+  const here = dirname(fileURLToPath(import.meta.url));
+  const pkg = JSON.parse(readFileSync2(join2(here, "..", "package.json"), "utf8"));
+  return pkg.version;
+}
+function printHelp() {
+  console.log(`slash-tokens — Token Optimization for Context Engineers
+
+  bunx slash-tokens          try (scan this directory, no account)
+  npm install slash-tokens   SDK
+
+  --version, -V    print version and exit
+  --help, -h       print this help and exit
+  --key=KEY        optional; CLI scans never charge
+
+Run in a project that already calls an LLM.
+Empty folder → no call sites. That's normal.`);
+}

package/dist/intercept.d.ts

@@ -11,6 +11,14 @@ export interface InterceptEvent {
     routed: boolean;
     timestamp: string;
 }
+/**
+ * The model a request actually names, for PRICING: the exact table entry when
+ * the strict canonical ID matches (claude-opus-5-5 → claude-opus-5.5,
+ * gpt-6-luna), else the legacy family mapping. Routing keeps using
+ * normalizeModel() so /auto decisions are unchanged within 1.6.x.
+ */
+export declare function identifyModel(raw: string): string;
 export declare function normalizeModel(raw: string): string;
+export declare function findCheapestRoute(provider: string, tokens: number, currentModel: string): string | null;
 export declare function onIntercept(handler: (event: InterceptEvent) => void): void;
 export declare function patchFetch(): void;

package/dist/intercept.js

@@ -1,20 +1,29 @@
 import { slash } from './slash.js';
-import { getModel } from './models.js';
+import { getModel, MODELS, canonicalModel } from './models.js';
 import { shouldRoute, isModelAllowed } from './config.js';
-import { PROVIDER_MODELS } from './providers.js';
+import { PROVIDER_MODELS, AUTO_ROUTE_TARGETS, NOT_ROUTE_TARGETS } from './providers.js';
 // Reverse lookup: model name → provider model names in the API
 // (what to put back in the request body)
 const MODEL_API_NAMES = {
+    'claude-opus-5': 'claude-opus-5',
     'claude-opus': 'claude-opus-5',
     'claude-opus-4.7': 'claude-opus-4-7',
+    'claude-sonnet-5': 'claude-sonnet-5',
     'claude-sonnet': 'claude-sonnet-5',
+    'claude-haiku-4.5': 'claude-haiku-4-5-20251001',
     'claude-haiku': 'claude-haiku-4-5-20251001',
+    'gpt-5.6-sol': 'gpt-5.6-sol',
+    'gpt-5.6-terra': 'gpt-5.6-terra',
+    'gpt-5.6-luna': 'gpt-5.6-luna',
     'gpt-5.4': 'gpt-5.4',
     'gpt-5.4-mini': 'gpt-5.4-mini',
     'gpt-5.4-nano': 'gpt-5.4-nano',
+    'grok-4.6': 'grok-4.6',
+    'grok-4.3': 'grok-4.3',
     'grok-4.20': 'grok-4.20-0309-non-reasoning',
     'grok-4-1-fast': 'grok-4.3',
     'gemini-3.1-pro': 'gemini-pro-latest',
+    'gemini-3.5-flash-lite': 'gemini-3.5-flash-lite',
     'gemini-2.5-flash': 'gemini-flash-latest',
 };
 // AI API endpoint detection
@@ -27,7 +36,7 @@ const AI_ENDPOINTS = [
     {
         pattern: /api\.openai\.com/,
         provider: 'OpenAI',
-        modelExtractor: (body) => body?.model || 'gpt-5.4',
+        modelExtractor: (body) => body?.model || 'gpt-5.6-sol',
     },
     {
         pattern: /generativelanguage\.googleapis\.com/,
@@ -35,13 +44,13 @@ const AI_ENDPOINTS = [
         modelExtractor: (_body, url) => {
             // Model is in the URL path: /v1beta/models/gemini-2.0-flash:generateContent
             const match = url?.match(/\/models\/([^/:]+)/);
-            return match ? match[1] : 'gemini-2.5-flash';
+            return match ? match[1] : 'gemini-3.5-flash-lite';
         },
     },
     {
         pattern: /api\.x\.ai/,
         provider: 'xAI',
-        modelExtractor: (body) => body?.model || 'grok-4.20',
+        modelExtractor: (body) => body?.model || 'grok-4.6',
     },
 ];
 // Normalize model names to our pricing table keys.
@@ -53,35 +62,66 @@ const AI_ENDPOINTS = [
 // (getModel() returning undefined → $0 reported cost, see slash.ts's
 // DEFAULT_UNKNOWN_MODEL_FACTOR comment for why "unrecognized" defaulting
 // to a falsely-safe-looking value is the dangerous case).
+/**
+ * The model a request actually names, for PRICING: the exact table entry when
+ * the strict canonical ID matches (claude-opus-5-5 → claude-opus-5.5,
+ * gpt-6-luna), else the legacy family mapping. Routing keeps using
+ * normalizeModel() so /auto decisions are unchanged within 1.6.x.
+ */
+export function identifyModel(raw) {
+    const c = canonicalModel(raw);
+    return MODELS[c] ? c : normalizeModel(raw);
+}
 export function normalizeModel(raw) {
     const lower = raw.toLowerCase();
-    // Anthropic — 4.7 before generic opus check, same order as slash-models.ts
+    // Anthropic — specific versions before generic family
     if (lower.includes('opus') && (lower.includes('4-7') || lower.includes('4.7')))
         return 'claude-opus-4.7';
+    if (lower.includes('opus-5'))
+        return 'claude-opus-5';
     if (lower.includes('opus'))
         return 'claude-opus';
+    if (lower.includes('sonnet-5'))
+        return 'claude-sonnet-5';
     if (lower.includes('sonnet'))
         return 'claude-sonnet';
+    if (lower.includes('haiku') && (lower.includes('4.5') || lower.includes('4-5')))
+        return 'claude-haiku-4.5';
     if (lower.includes('haiku'))
         return 'claude-haiku';
-    // xAI
+    // xAI — 4.6 / 4.3 / 4.20 before generic grok
+    if (lower.includes('grok') && (lower.includes('4.6') || lower.includes('4-6')))
+        return 'grok-4.6';
+    if (lower.includes('grok') && (lower.includes('4.3') || lower.includes('4-3')))
+        return 'grok-4.3';
+    if (lower.includes('grok') && (lower.includes('4.20') || lower.includes('4-20')))
+        return 'grok-4.20';
     if (lower.includes('grok') && lower.includes('fast'))
         return 'grok-4-1-fast';
     if (lower.includes('grok'))
-        return 'grok-4.20';
+        return 'grok-4.6';
     // Google
     if (lower.includes('gemini') && lower.includes('pro'))
         return 'gemini-3.1-pro';
-    if (lower.includes('gemini'))
+    if (lower.includes('gemini') && lower.includes('3.5') && lower.includes('lite'))
+        return 'gemini-3.5-flash-lite';
+    if (lower.includes('gemini') && lower.includes('2.5'))
         return 'gemini-2.5-flash';
-    // OpenAI — current (5.4 family)
+    if (lower.includes('gemini'))
+        return 'gemini-3.5-flash-lite';
+    // OpenAI — 5.6 then 5.4 then legacy
+    if (lower.includes('5.6') && lower.includes('luna'))
+        return 'gpt-5.6-luna';
+    if (lower.includes('5.6') && lower.includes('terra'))
+        return 'gpt-5.6-terra';
+    if (lower.includes('5.6'))
+        return 'gpt-5.6-sol';
     if (lower.includes('5.4') && lower.includes('nano'))
         return 'gpt-5.4-nano';
     if (lower.includes('5.4') && lower.includes('mini'))
         return 'gpt-5.4-mini';
     if (lower.includes('5.4'))
         return 'gpt-5.4';
-    // OpenAI — legacy model names → map to closest current equivalent
     if (lower.includes('o1-mini') || lower.includes('o1_mini'))
         return 'gpt-5.4-mini';
     if (lower.includes('o1'))
@@ -118,7 +158,7 @@ function extractContent(body) {
     return JSON.stringify(body);
 }
 // Find cheapest model from same provider that fits
-function findCheapestRoute(provider, tokens, currentModel) {
+export function findCheapestRoute(provider, tokens, currentModel) {
     const providerModels = PROVIDER_MODELS[provider];
     if (!providerModels)
         return null;
@@ -126,6 +166,10 @@ function findCheapestRoute(provider, tokens, currentModel) {
     for (const model of providerModels) {
         if (model === currentModel)
             continue;
+        if (!AUTO_ROUTE_TARGETS.has(model))
+            continue; // 1.6.x: /auto rewrites only to its 1.6.5 targets
+        if (NOT_ROUTE_TARGETS.has(model))
+            continue; // specialised models are never targets
         if (!isModelAllowed(model))
             continue; // user excluded this model
         const info = getModel(model);
@@ -158,13 +202,14 @@ export function patchFetch() {
                     const body = JSON.parse(bodyStr);
                     const content = extractContent(body);
                     const rawModel = match.modelExtractor(body, url);
-                    const originalModel = normalizeModel(rawModel);
+                    const routingModel = normalizeModel(rawModel); // decides routes (unchanged in 1.6.x)
+                    const originalModel = identifyModel(rawModel); // prices the request it really names
                     const tokens = slash(content, originalModel);
                     const originalInfo = getModel(originalModel);
                     const originalCost = originalInfo ? Math.round(((tokens / 1000000) * originalInfo.input) * 1000000) / 1000000 : 0;
                     const fits = originalInfo ? tokens <= originalInfo.context : true;
                     // Find cheapest route within same provider (if routing enabled)
-                    const routeModel = shouldRoute() ? findCheapestRoute(match.provider, tokens, originalModel) : null;
+                    const routeModel = shouldRoute() ? findCheapestRoute(match.provider, tokens, routingModel) : null;
                     const routedInfo = routeModel ? getModel(routeModel) : null;
                     const routedCost = routedInfo ? Math.round(((tokens / 1000000) * routedInfo.input) * 1000000) / 1000000 : originalCost;
                     const salvaged = routeModel ? Math.round((originalCost - routedCost) * 1000000) / 1000000 : 0;

package/dist/models.d.ts

@@ -7,6 +7,13 @@ export interface ModelInfo {
     longContextOutput?: number;
 }
 export declare const MODELS: Record<string, ModelInfo>;
+/**
+ * Real API IDs → table keys, strictly: lowercase, drop a trailing date stamp
+ * (`-20250514`), and write version numbers with dots (`claude-opus-4-7` →
+ * `claude-opus-4.7`). No family guessing: an unknown version stays unknown,
+ * so a new model never silently gets an older model's price.
+ */
+export declare function canonicalModel(name: string): string;
 export declare function getModel(name: string): ModelInfo | undefined;
 export declare function effectiveRate(tokens: number, info: ModelInfo): {
     input: number;

package/dist/models.js

@@ -1,44 +1,119 @@
-// Pricing as of April 2026 — USD per million tokens
-// Claude Sonnet 5 pricing corrected 2026-08-23: was hardcoded at $3.00/
-// $15.00 (the previously-scheduled Sept 1, 2026 increase), but Anthropic's
-// own pricing page confirms that increase will NOT occur — $2.00/$10.00
-// (the introductory rate) is now the permanent standard price. Found via
-// a full-codebase review that independently re-verified every provider's
-// live pricing, not just the model IDs already fixed that day.
-// xAI pricing re-derived 2026-08-23: grok-4.20 and grok-4-1-fast (the literal
-// API IDs, not just the generic keys below) were both fully retired — not
-// just old snapshots, they 404 on the live API. Current lineup has no cheap
-// tier at all; grok-4.20 (generic) now targets grok-4.20-0309-non-reasoning
-// and grok-4-1-fast (generic) targets grok-4.3, both $1.25/$2.50/1M — see
-// intercept.ts MODEL_API_NAMES. There is currently no xAI model cheaper than
-// $1.25/M input, so routing between these two generic keys yields zero
-// savings (findCheapestRoute requires strictly cheaper — this is honest,
-// not a bug: the old $0.20/M "fast" tier no longer exists).
+const OPUS = { input: 5.00, output: 25.00, context: 1000000 };
+const OPUS_55 = { input: 4.00, output: 20.00, context: 1000000 };
+const FABLE = { input: 10.00, output: 50.00, context: 1000000 };
+const SONNET_4X = { input: 3.00, output: 15.00, context: 1000000 };
+const SONNET = { input: 2.00, output: 10.00, context: 1000000 };
+const HAIKU = { input: 1.00, output: 5.00, context: 200000 };
+const GROK_46 = {
+    input: 2.00, output: 6.00, context: 500000,
+    longContextThreshold: 200000, longContextInput: 4.00, longContextOutput: 12.00,
+};
+const GROK_43 = {
+    input: 1.25, output: 2.50, context: 1000000,
+    longContextThreshold: 200000, longContextInput: 2.50, longContextOutput: 5.00,
+};
+const GEMINI_PRO = {
+    input: 2.00, output: 12.00, context: 1000000,
+    longContextThreshold: 200000, longContextInput: 4.00, longContextOutput: 18.00,
+};
+const GEMINI_FLASH = { input: 0.30, output: 2.50, context: 1000000 };
+// Gemini 3.6–3.8 Flash: launch price through 2026-12-31; Google lists $1.50/$7.50
+// from 2027-01-01. Update this entry before then.
+const GEMINI_FLASH_3X = { input: 0.75, output: 3.75, context: 1000000 };
+const GEMINI_35_FLASH = { input: 1.50, output: 9.00, context: 1000000 };
+const GEMINI_31_FLASH_LITE = { input: 0.25, output: 1.50, context: 1000000 };
+const GROK_BUILD = {
+    input: 1.00, output: 2.00, context: 256000,
+    longContextThreshold: 200000, longContextInput: 2.00, longContextOutput: 4.00,
+};
+const GPT_6_ASTRA = { input: 10.00, output: 50.00, context: 1050000 };
+const GPT_6_SOL = { input: 2.00, output: 10.00, context: 1050000 };
+const GPT_6_LUNA = { input: 0.10, output: 0.50, context: 1050000 };
+const GPT_SOL = { input: 4.00, output: 20.00, context: 1050000 };
+const GPT_TERRA = { input: 2.00, output: 12.00, context: 1050000 };
+const GPT_LUNA = { input: 0.20, output: 1.20, context: 1050000 };
+const GPT_54 = { input: 2.50, output: 15.00, context: 1000000 };
+const GPT_54_MINI = { input: 0.75, output: 4.50, context: 128000 };
+const GPT_54_NANO = { input: 0.20, output: 1.25, context: 128000 };
+// Pricing as of 2026-10-07 — USD per million tokens (checked against the four pages below).
+// First-party: platform.claude.com/docs/en/about-claude/pricing
+//              developers.openai.com/api/docs/models
+//              docs.x.ai/developers/models
+//              ai.google.dev/gemini-api/docs/pricing
+// Old keys stay as aliases so existing call sites don't throw.
 export const MODELS = {
-    // Anthropic
-    'claude-opus': { input: 5.00, output: 25.00, context: 1000000 },
-    'claude-opus-4.7': { input: 5.00, output: 25.00, context: 1000000 },
-    'claude-sonnet': { input: 2.00, output: 10.00, context: 1000000 },
-    'claude-haiku': { input: 1.00, output: 5.00, context: 200000 },
-    // xAI
-    'grok-4.20': { input: 1.25, output: 2.50, context: 1000000, longContextThreshold: 200000, longContextInput: 2.50, longContextOutput: 5.00 },
-    'grok-4-1-fast': { input: 1.25, output: 2.50, context: 1000000, longContextThreshold: 200000, longContextInput: 2.50, longContextOutput: 5.00 },
+    // Anthropic — live names + generic aliases (same rates)
+    'claude-fable-5.1': { ...FABLE },
+    'claude-fable-5': { ...FABLE },
+    'claude-mythos-5.1': { ...FABLE },
+    'claude-mythos-5': { ...FABLE },
+    'claude-opus-5.5': { ...OPUS_55 },
+    'claude-opus-5': { ...OPUS },
+    'claude-opus-4.8': { ...OPUS },
+    'claude-opus': { ...OPUS },
+    'claude-opus-4.7': { ...OPUS },
+    'claude-opus-4.6': { ...OPUS },
+    'claude-opus-4.5': { ...OPUS },
+    'claude-sonnet-5.5': { ...SONNET },
+    'claude-sonnet-5': { ...SONNET },
+    'claude-sonnet': { ...SONNET },
+    'claude-sonnet-4.6': { ...SONNET_4X },
+    'claude-sonnet-4.5': { ...SONNET_4X },
+    'claude-haiku-4.5': { ...HAIKU },
+    'claude-haiku': { ...HAIKU },
+    // xAI — flagship 4.6, cheap same-provider 4.3. 4.20 / fast are aliases.
+    'grok-4.7': { ...GROK_46 },
+    'grok-4.6': { ...GROK_46 },
+    'grok-4.5': { ...GROK_46 },
+    'grok-build-0.1': { ...GROK_BUILD },
+    'grok-4.3': { ...GROK_43 },
+    'grok-4.20': { ...GROK_43 },
+    'grok-4-1-fast': { ...GROK_43 },
     // Google
-    'gemini-3.1-pro': { input: 2.00, output: 12.00, context: 1000000 },
-    'gemini-2.5-flash': { input: 0.30, output: 2.50, context: 1000000 },
-    // OpenAI
-    'gpt-5.4': { input: 2.50, output: 15.00, context: 1000000 },
-    'gpt-5.4-mini': { input: 0.75, output: 4.50, context: 128000 },
-    'gpt-5.4-nano': { input: 0.20, output: 1.25, context: 128000 },
+    'gemini-3.1-pro': { ...GEMINI_PRO },
+    'gemini-3.1-pro-preview': { ...GEMINI_PRO },
+    'gemini-3.8-flash': { ...GEMINI_FLASH_3X },
+    'gemini-3.7-flash': { ...GEMINI_FLASH_3X },
+    'gemini-3.6-flash': { ...GEMINI_FLASH_3X },
+    'gemini-3.5-flash': { ...GEMINI_35_FLASH },
+    'gemini-3.1-flash-lite': { ...GEMINI_31_FLASH_LITE },
+    'gemini-3.5-flash-lite': { ...GEMINI_FLASH },
+    'gemini-2.5-flash': { ...GEMINI_FLASH },
+    // OpenAI — live 5.6 ladder. 5.4 family kept as aliases (old prices).
+    'gpt-6-astra': { ...GPT_6_ASTRA },
+    'gpt-6.1-sol': { ...GPT_6_SOL },
+    'gpt-6-sol': { ...GPT_6_SOL },
+    'gpt-6-luna': { ...GPT_6_LUNA },
+    'gpt-5.6-sol': { ...GPT_SOL },
+    'gpt-5.6-terra': { ...GPT_TERRA },
+    'gpt-5.6-luna': { ...GPT_LUNA },
+    'gpt-5.4': { ...GPT_54 },
+    'gpt-5.4-mini': { ...GPT_54_MINI },
+    'gpt-5.4-nano': { ...GPT_54_NANO },
 };
+/**
+ * Real API IDs → table keys, strictly: lowercase, drop a trailing date stamp
+ * (`-20250514`), and write version numbers with dots (`claude-opus-4-7` →
+ * `claude-opus-4.7`). No family guessing: an unknown version stays unknown,
+ * so a new model never silently gets an older model's price.
+ */
+export function canonicalModel(name) {
+    if (MODELS[name])
+        return name;
+    let n = name.trim().toLowerCase();
+    if (MODELS[n])
+        return n;
+    n = n.replace(/-\d{8}$/, '');
+    if (MODELS[n])
+        return n;
+    const dotted = n.replace(/(\d)-(\d)(?=$|-)/g, '$1.$2');
+    if (MODELS[dotted])
+        return dotted;
+    return n;
+}
 export function getModel(name) {
-    return MODELS[name] || MODELS[name.toLowerCase()];
+    return MODELS[canonicalModel(name)];
 }
-// Resolve the actual billable rate for a given token count — applies the
-// long-context tier above if the model has one and tokens cross it.
-// preflight()/preflightRoute() should go through this, not read
-// .input/.output directly, or a Grok call over 200K tokens gets silently
-// under-costed at the base rate.
 export function effectiveRate(tokens, info) {
     if (info.longContextThreshold !== undefined && tokens > info.longContextThreshold) {
         return {

package/dist/patterns.d.ts

@@ -4,6 +4,6 @@ export interface Pattern {
 }
 export declare const AI_PATTERNS: Pattern[];
 export declare const SDK_REPRESENTATIVE_MODEL: Record<string, string>;
-export declare const UNKNOWN_SDK_REPRESENTATIVE_MODEL = "claude-sonnet";
+export declare const UNKNOWN_SDK_REPRESENTATIVE_MODEL = "claude-sonnet-5";
 export declare const SKIP_DIRS: Set<string>;
 export declare const SCAN_EXTENSIONS: Set<string>;

package/dist/patterns.js

@@ -34,17 +34,17 @@ export const AI_PATTERNS = [
 // know the exact model), but a real per-provider one instead of a single
 // guess applied to everyone.
 export const SDK_REPRESENTATIVE_MODEL = {
-    'Anthropic': 'claude-sonnet',
-    'OpenAI': 'gpt-5.4',
+    'Anthropic': 'claude-sonnet-5',
+    'OpenAI': 'gpt-5.6-sol',
     'Gemini': 'gemini-3.1-pro',
-    'Grok': 'grok-4.20',
+    'Grok': 'grok-4.6',
 };
 // Fallback for SDKs that don't map to one specific provider (Vercel AI,
 // LangChain, and Bedrock can all wrap any underlying provider; raw
 // fetch-to-AI-endpoint and Cohere/Mistral have no pricing data in MODELS
 // at all). claude-sonnet is used as a documented, honest middle-of-the-
 // road placeholder — not a claim about which model is actually running.
-export const UNKNOWN_SDK_REPRESENTATIVE_MODEL = 'claude-sonnet';
+export const UNKNOWN_SDK_REPRESENTATIVE_MODEL = 'claude-sonnet-5';
 export const SKIP_DIRS = new Set([
     'node_modules', '.git', 'dist', 'build', '.next', '.nuxt', '.svelte-kit',
     'coverage', '.turbo', '.cache', '__pycache__', '.venv', 'venv',

package/dist/preflight.js

@@ -1,6 +1,6 @@
 import { slash } from './slash.js';
-import { getModel, MODELS, effectiveRate } from './models.js';
-import { PROVIDER_MODELS, providerOf } from './providers.js';
+import { getModel, MODELS, effectiveRate, canonicalModel } from './models.js';
+import { PROVIDER_MODELS, providerOf, NOT_ROUTE_TARGETS } from './providers.js';
 import { shouldRoute, isModelAllowed } from './config.js';
 /**
  * Compute cost of a prompt of `tokens` tokens on the given ModelInfo.
@@ -99,9 +99,12 @@ export function preflightRoute(content, model) {
         return null;
     const originalCost = computeCost(tokens, info);
     let cheapest = null;
+    const self = canonicalModel(model);
     for (const m of providerModels) {
-        if (m === model)
+        if (m === self)
             continue;
+        if (NOT_ROUTE_TARGETS.has(m))
+            continue; // specialised: never a target
         if (!isModelAllowed(m))
             continue; // user excluded this model
         const altInfo = getModel(m);
@@ -112,7 +115,10 @@ export function preflightRoute(content, model) {
         if (altInfo.input >= info.input)
             continue; // not cheaper
         const alt = buildAlternative(m, originalCost, tokens, altInfo);
-        if (!cheapest || alt.cost < cheapest.cost) {
+        // Tiny prompts round costs to the same value; break ties on the list price
+        // so the genuinely cheaper model wins (gpt-5.6-sol → gpt-6-luna, not 5.4-nano).
+        if (!cheapest || alt.cost < cheapest.cost ||
+            (alt.cost === cheapest.cost && altInfo.input < (getModel(cheapest.model)?.input ?? Infinity))) {
             cheapest = alt;
         }
     }

package/dist/providers.d.ts

@@ -1,26 +1,24 @@
 /**
  * Provider groups — single source of truth.
  *
- * Slash routing is always SAME-PROVIDER. Opus → Haiku, GPT-5.4 → Nano,
- * Grok-4.20 → Fast, Gemini 3.1 Pro → 2.5 Flash. Never cross-provider.
+ * Slash routing is always SAME-PROVIDER. Opus 5 → Haiku 4.5, Sol → Luna,
+ * Grok 4.6 → 4.3, Gemini 3.1 Pro → 3.5 Flash-Lite. Never cross-provider.
  *
- * This file is the canonical provider mapping. Both `intercept.ts`
- * (runtime fetch patching) and `preflight.ts` (analysis + routing
- * prediction) consume from here so they can never drift.
- *
- * TEST-NOTE: Whenever a new model is added, it MUST appear in exactly one
- * provider group below. A model missing from here will:
- *   - Never be a routing target from `findCheapestRoute` / `preflightRoute`
- *   - Still appear in `preflight().options` (which is cross-provider analysis)
- *   That mismatch is by design — see preflight.ts semantics.
+ * Order inside a group matters when two models share a price: the first
+ * strictly-cheaper hit wins (findCheapestRoute / preflightRoute).
  */
 export declare const PROVIDER_MODELS: Record<string, string[]>;
 /**
- * Identify a model's provider from its canonical key.
- *
- * TEST-NOTE: This function MUST return a non-null provider for every model
- * present in `MODELS` (from models.ts). If `MODELS` adds a model without
- * adding it to `PROVIDER_MODELS`, this returns null and routing is disabled
- * for that model. Silent skip.
+ * Priced and grouped, but never a routing TARGET: specialised models a general
+ * call shouldn't be moved to (grok-build-0.1 is a coding-agent model).
+ */
+export declare const NOT_ROUTE_TARGETS: ReadonlySet<string>;
+/**
+ * Models `slash-tokens/auto` may rewrite a live request TO. Frozen at the
+ * 1.6.5 set so a patch release never changes where production calls go:
+ * models added in 1.6.6 (GPT-6, Gemini 3.1 Flash-Lite / 3.6–3.8 Flash, Grok
+ * 4.7 / 4.5, Claude 5.5 / Fable …) are priced and shown by preflight() /
+ * preflightRoute(), but /auto only starts routing to them in 1.7.0.
  */
+export declare const AUTO_ROUTE_TARGETS: ReadonlySet<string>;
 export declare function providerOf(model: string): string | null;

package/dist/providers.js

@@ -1,36 +1,58 @@
+import { canonicalModel } from './models.js';
 /**
  * Provider groups — single source of truth.
  *
- * Slash routing is always SAME-PROVIDER. Opus → Haiku, GPT-5.4 → Nano,
- * Grok-4.20 → Fast, Gemini 3.1 Pro → 2.5 Flash. Never cross-provider.
+ * Slash routing is always SAME-PROVIDER. Opus 5 → Haiku 4.5, Sol → Luna,
+ * Grok 4.6 → 4.3, Gemini 3.1 Pro → 3.5 Flash-Lite. Never cross-provider.
  *
- * This file is the canonical provider mapping. Both `intercept.ts`
- * (runtime fetch patching) and `preflight.ts` (analysis + routing
- * prediction) consume from here so they can never drift.
- *
- * TEST-NOTE: Whenever a new model is added, it MUST appear in exactly one
- * provider group below. A model missing from here will:
- *   - Never be a routing target from `findCheapestRoute` / `preflightRoute`
- *   - Still appear in `preflight().options` (which is cross-provider analysis)
- *   That mismatch is by design — see preflight.ts semantics.
+ * Order inside a group matters when two models share a price: the first
+ * strictly-cheaper hit wins (findCheapestRoute / preflightRoute).
  */
 export const PROVIDER_MODELS = {
-    Anthropic: ['claude-opus', 'claude-opus-4.7', 'claude-sonnet', 'claude-haiku'],
-    OpenAI: ['gpt-5.4', 'gpt-5.4-mini', 'gpt-5.4-nano'],
-    xAI: ['grok-4.20', 'grok-4-1-fast'],
-    Google: ['gemini-3.1-pro', 'gemini-2.5-flash'],
+    Anthropic: [
+        'claude-fable-5.1', 'claude-fable-5', 'claude-mythos-5.1', 'claude-mythos-5',
+        'claude-opus-5', 'claude-opus', 'claude-opus-4.8', 'claude-opus-4.7',
+        'claude-opus-4.6', 'claude-opus-4.5', 'claude-opus-5.5',
+        'claude-sonnet-4.6', 'claude-sonnet-4.5',
+        'claude-sonnet-5.5', 'claude-sonnet-5', 'claude-sonnet',
+        'claude-haiku', 'claude-haiku-4.5',
+    ],
+    OpenAI: [
+        'gpt-6-astra',
+        'gpt-5.6-sol', 'gpt-5.6-terra',
+        'gpt-5.4', 'gpt-6.1-sol', 'gpt-6-sol', 'gpt-5.4-mini', 'gpt-5.4-nano',
+        'gpt-5.6-luna', 'gpt-6-luna',
+    ],
+    xAI: ['grok-4.7', 'grok-4.6', 'grok-4.5', 'grok-4.3', 'grok-4.20', 'grok-4-1-fast', 'grok-build-0.1'],
+    Google: [
+        'gemini-3.1-pro', 'gemini-3.1-pro-preview',
+        'gemini-3.5-flash', 'gemini-3.8-flash', 'gemini-3.7-flash', 'gemini-3.6-flash',
+        'gemini-3.5-flash-lite', 'gemini-2.5-flash', 'gemini-3.1-flash-lite',
+    ],
 };
 /**
- * Identify a model's provider from its canonical key.
- *
- * TEST-NOTE: This function MUST return a non-null provider for every model
- * present in `MODELS` (from models.ts). If `MODELS` adds a model without
- * adding it to `PROVIDER_MODELS`, this returns null and routing is disabled
- * for that model. Silent skip.
+ * Priced and grouped, but never a routing TARGET: specialised models a general
+ * call shouldn't be moved to (grok-build-0.1 is a coding-agent model).
+ */
+export const NOT_ROUTE_TARGETS = new Set(['grok-build-0.1']);
+/**
+ * Models `slash-tokens/auto` may rewrite a live request TO. Frozen at the
+ * 1.6.5 set so a patch release never changes where production calls go:
+ * models added in 1.6.6 (GPT-6, Gemini 3.1 Flash-Lite / 3.6–3.8 Flash, Grok
+ * 4.7 / 4.5, Claude 5.5 / Fable …) are priced and shown by preflight() /
+ * preflightRoute(), but /auto only starts routing to them in 1.7.0.
  */
+export const AUTO_ROUTE_TARGETS = new Set([
+    'claude-opus-5', 'claude-opus', 'claude-opus-4.7', 'claude-sonnet-5', 'claude-sonnet',
+    'claude-haiku', 'claude-haiku-4.5',
+    'gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.4', 'gpt-5.4-mini', 'gpt-5.4-nano', 'gpt-5.6-luna',
+    'grok-4.6', 'grok-4.3', 'grok-4.20', 'grok-4-1-fast',
+    'gemini-3.1-pro', 'gemini-3.5-flash-lite', 'gemini-2.5-flash',
+]);
 export function providerOf(model) {
+    const id = canonicalModel(model);
     for (const [provider, models] of Object.entries(PROVIDER_MODELS)) {
-        if (models.includes(model))
+        if (models.includes(id))
             return provider;
     }
     return null;

package/dist/report.js

@@ -24,6 +24,7 @@ export function printReport(sites, filesScanned, timeMs, cwd) {
     if (sites.length === 0) {
         console.log(`${WHITE}  No AI API call sites detected.${R}`);
         console.log(`${GRAY}  Supported: OpenAI, Anthropic, Vercel AI, LangChain, Gemini, Bedrock, Grok${R}`);
+        console.log(`${GRAY}  Run this in a project that already calls an LLM.${R}`);
         console.log('');
         return;
     }

package/dist/slash.js

@@ -1,4 +1,5 @@
 import { getInstance, writeToMemory, ensureCapacity } from './wasm.js';
+import { canonicalModel } from './models.js';
 const WASM_INPUT_OFFSET = 4096;
 /**
  * Per-model calibration factors.
@@ -65,7 +66,10 @@ const WASM_INPUT_OFFSET = 4096;
  *             DEFAULT_UNKNOWN_MODEL_FACTOR (1.85), a ~60% larger correction
  *             than it needed. Still safe either way (1.85 > required
  *             minimum), just needlessly inflated for real GPT users.
- * grok-4.20 / grok-4-1-fast: 1.15 — re-verified 2026-08-23 against the
+ * grok-4.6 / grok-4.3 (and aliases grok-4.20 / grok-4-1-fast): 1.15 —
+ *             4.6/4.3 IDs added 2026-08-25; factor CARRIED from the
+ *             2026-08-23 corpus, not re-measured on the new wire IDs.
+ * grok-4.20 / grok-4-1-fast (original): 1.15 — re-verified 2026-08-23 against the
  *             29-sample corpus (20 new samples: more languages, Spanish/
  *             Japanese prose, more JSON shapes) and UNCHANGED — same
  *             worst case (technical-docs prose, ratio 0.928) as the
@@ -91,14 +95,43 @@ const WASM_INPUT_OFFSET = 4096;
  * Slash must NEVER under-report. Over-reporting is safe (go/no-go only).
  */
 const CALIBRATION = {
+    'claude-fable-5.1': 2.05,
+    'claude-fable-5': 2.05,
+    'claude-mythos-5.1': 2.05,
+    'claude-mythos-5': 2.05,
+    'claude-opus-5.5': 2.05,
+    'claude-opus-5': 2.05,
+    'claude-opus-4.8': 2.05,
+    'claude-opus-4.6': 2.05,
+    'claude-opus-4.5': 2.05,
+    'claude-sonnet-5.5': 2.05,
+    'claude-sonnet-4.6': 2.05,
+    'claude-sonnet-4.5': 2.05,
     'claude-opus': 2.05,
     'claude-opus-4.7': 2.05,
+    'claude-sonnet-5': 2.05,
     'claude-sonnet': 2.05,
+    'claude-haiku-4.5': 1.45,
     'claude-haiku': 1.45,
     'gemini-3.1-pro': 1.45,
+    'gemini-3.1-pro-preview': 1.45,
+    'gemini-3.8-flash': 1.45,
+    'gemini-3.7-flash': 1.45,
+    'gemini-3.6-flash': 1.45,
+    'gemini-3.5-flash': 1.45,
+    'gemini-3.1-flash-lite': 1.45,
+    'gemini-3.5-flash-lite': 1.45,
     'gemini-2.5-flash': 1.45,
+    'grok-4.7': 1.15,
+    'grok-4.6': 1.15,
+    'grok-4.5': 1.15,
+    'grok-build-0.1': 1.15,
+    'grok-4.3': 1.15,
     'grok-4.20': 1.15,
     'grok-4-1-fast': 1.15,
+    'gpt-5.6-sol': 1.15,
+    'gpt-5.6-terra': 1.15,
+    'gpt-5.6-luna': 1.15,
     'gpt-5.4': 1.15,
     'gpt-5.4-mini': 1.15,
     'gpt-5.4-nano': 1.15,
@@ -129,7 +162,10 @@ export function slash(content, model) {
     const raw = instance.exports.estimate_tokens(WASM_INPUT_OFFSET, len);
     if (!model)
         return raw;
-    const factor = CALIBRATION[model] ?? DEFAULT_UNKNOWN_MODEL_FACTOR;
+    // GPT-6 (gpt-6-astra / 6.1-sol / 6-sol / 6-luna) is deliberately NOT in
+    // CALIBRATION: a new generation with an unbenchmarked tokenizer takes the
+    // conservative default (never under-report) until a bench run adds it.
+    const factor = CALIBRATION[canonicalModel(model)] ?? DEFAULT_UNKNOWN_MODEL_FACTOR;
     return factor === 1.0 ? raw : Math.ceil(raw * factor);
 }
 /**

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "slash-tokens",
-  "version": "1.6.4",
+  "version": "1.6.6",
   "description": "Token Optimization for Context Engineers. 4.8 KB WASM. Sub-millisecond. Zero dependencies.",
   "main": "dist/index.js",
   "module": "dist/index.js",
@@ -26,8 +26,10 @@
   ],
   "sideEffects": false,
   "scripts": {
+    "prepublishOnly": "npm run build",
     "build": "npx tsc && bun build src/cli.ts --outfile=dist/cli.js --target=node",
-    "test": "bun test"
+    "test": "bun test",
+    "check:engines": "node scripts/check-engines.mjs"
   },
   "author": "wolfejam",
   "license": "MIT",
@@ -54,5 +56,8 @@
     "@types/node": "^25.5.0",
     "js-tiktoken": "^1.0.21",
     "typescript": "^6.0.2"
+  },
+  "engines": {
+    "node": ">=22.0.0"
   }
 }