@stll/ai-catalog 0.4.0 → 0.6.0

package/dist/benchmarks.js

@@ -0,0 +1,613 @@
+import { D as getModelRate, G as getStandardModelRate, W as MODEL_RATE_UNITS_PER_USD } from "./src-DaH9EEfe.js";
+//#region src/benchmark-sources.ts
+const MODEL_BENCHMARK_NAME = "Text Arena";
+const MODEL_BENCHMARK_LICENCE = "CC BY 4.0";
+const MODEL_BENCHMARK_DATASET = "lmarena-ai/leaderboard-dataset";
+const MODEL_BENCHMARK_SOURCE_URL = `https://huggingface.co/datasets/${MODEL_BENCHMARK_DATASET}`;
+/** Style-controlled ratings discount formatting and length preferences. */
+const MODEL_BENCHMARK_CONFIG = "text_style_control";
+const MODEL_BENCHMARK_SPLIT = "latest";
+const MODEL_BENCHMARK_CATEGORY = "overall";
+const MODEL_BENCHMARK_SOURCES = {
+	"gemini-3.8-flash": [{
+		sourceModelId: "gemini-3.8-flash-high",
+		reasoningEffort: "high"
+	}],
+	"gemini-3.7-flash": [{
+		sourceModelId: "gemini-3.7-flash-high",
+		reasoningEffort: "high"
+	}],
+	"gemini-3.6-flash": [{
+		sourceModelId: "gemini-3.6-flash-high",
+		reasoningEffort: "high"
+	}],
+	"gemini-3.5-flash-lite": [{
+		sourceModelId: "gemini-3.5-flash-lite",
+		reasoningEffort: null
+	}],
+	"gemini-3.1-pro-preview": [{
+		sourceModelId: "gemini-3.1-pro-preview",
+		reasoningEffort: null
+	}],
+	"gemini-3.5-flash": [{
+		sourceModelId: "gemini-3.5-flash-medium",
+		reasoningEffort: "medium"
+	}, {
+		sourceModelId: "gemini-3.5-flash-high",
+		reasoningEffort: "high"
+	}],
+	"gemini-3.1-flash-lite": [],
+	"openai/gpt-6.1-sol": [],
+	"openai/gpt-6-astra": [{
+		sourceModelId: "gpt-6-astra-max",
+		reasoningEffort: "max"
+	}],
+	"openai/gpt-6-sol": [{
+		sourceModelId: "gpt-6-sol-max",
+		reasoningEffort: "max"
+	}],
+	"openai/gpt-6-luna": [{
+		sourceModelId: "gpt-6-luna-max",
+		reasoningEffort: "max"
+	}],
+	"openai/gpt-5.6-sol": [{
+		sourceModelId: "gpt-5.6-sol-xhigh",
+		reasoningEffort: "xhigh"
+	}],
+	"openai/gpt-5.6-terra": [{
+		sourceModelId: "gpt-5.6-terra-xhigh",
+		reasoningEffort: "xhigh"
+	}],
+	"openai/gpt-5.6-luna": [{
+		sourceModelId: "gpt-5.6-luna-xhigh",
+		reasoningEffort: "xhigh"
+	}],
+	"google/gemini-3.8-flash": [{
+		sourceModelId: "gemini-3.8-flash-high",
+		reasoningEffort: "high"
+	}],
+	"google/gemini-3.7-flash": [{
+		sourceModelId: "gemini-3.7-flash-high",
+		reasoningEffort: "high"
+	}],
+	"google/gemini-3.6-flash": [{
+		sourceModelId: "gemini-3.6-flash-high",
+		reasoningEffort: "high"
+	}],
+	"google/gemini-3.5-flash-lite": [],
+	"google/gemini-3.1-pro-preview": [],
+	"google/gemini-3.5-flash": [{
+		sourceModelId: "gemini-3.5-flash-medium",
+		reasoningEffort: "medium"
+	}, {
+		sourceModelId: "gemini-3.5-flash-high",
+		reasoningEffort: "high"
+	}],
+	"google/gemini-3.1-flash-lite": [],
+	"anthropic/claude-sonnet-5.5": [],
+	"anthropic/claude-sonnet-5": [{
+		sourceModelId: "claude-sonnet-5-high",
+		reasoningEffort: "high"
+	}],
+	"anthropic/claude-opus-5": [{
+		sourceModelId: "claude-opus-5-high",
+		reasoningEffort: "high"
+	}, {
+		sourceModelId: "claude-opus-5-max",
+		reasoningEffort: "max"
+	}],
+	"anthropic/claude-opus-4.8": [{
+		sourceModelId: "claude-opus-4-8-high",
+		reasoningEffort: "high"
+	}],
+	"anthropic/claude-sonnet-4.6": [],
+	"openai/gpt-5.5": [{
+		sourceModelId: "gpt-5.5-high",
+		reasoningEffort: "high"
+	}],
+	"openai/gpt-5.4-mini": [{
+		sourceModelId: "gpt-5.4-mini-high",
+		reasoningEffort: "high"
+	}],
+	"gpt-6.1-sol": [],
+	"gpt-6-astra": [{
+		sourceModelId: "gpt-6-astra-max",
+		reasoningEffort: "max"
+	}],
+	"gpt-6-sol": [{
+		sourceModelId: "gpt-6-sol-max",
+		reasoningEffort: "max"
+	}],
+	"gpt-6-luna": [{
+		sourceModelId: "gpt-6-luna-max",
+		reasoningEffort: "max"
+	}],
+	"gpt-5.6": [{
+		sourceModelId: "gpt-5.6-sol-xhigh",
+		reasoningEffort: "xhigh"
+	}],
+	"gpt-5.6-terra": [{
+		sourceModelId: "gpt-5.6-terra-xhigh",
+		reasoningEffort: "xhigh"
+	}],
+	"gpt-5.6-luna": [{
+		sourceModelId: "gpt-5.6-luna-xhigh",
+		reasoningEffort: "xhigh"
+	}],
+	"gpt-5.5": [{
+		sourceModelId: "gpt-5.5",
+		reasoningEffort: null
+	}, {
+		sourceModelId: "gpt-5.5-high",
+		reasoningEffort: "high"
+	}],
+	"gpt-5.4": [{
+		sourceModelId: "gpt-5.4",
+		reasoningEffort: null
+	}, {
+		sourceModelId: "gpt-5.4-high",
+		reasoningEffort: "high"
+	}],
+	"gpt-5.4-mini": [{
+		sourceModelId: "gpt-5.4-mini-high",
+		reasoningEffort: "high"
+	}],
+	"gpt-5.4-nano": [{
+		sourceModelId: "gpt-5.4-nano-high",
+		reasoningEffort: "high"
+	}],
+	"gpt-5.2": [{
+		sourceModelId: "gpt-5.2",
+		reasoningEffort: null
+	}, {
+		sourceModelId: "gpt-5.2-high",
+		reasoningEffort: "high"
+	}],
+	"claude-sonnet-5-5": [],
+	"claude-sonnet-5": [{
+		sourceModelId: "claude-sonnet-5-high",
+		reasoningEffort: "high"
+	}],
+	"claude-fable-5-1": [{
+		sourceModelId: "claude-fable-5.1-max",
+		reasoningEffort: "max"
+	}],
+	"claude-fable-5": [{
+		sourceModelId: "claude-fable-5-high",
+		reasoningEffort: "high"
+	}],
+	"claude-opus-5-5": [{
+		sourceModelId: "claude-opus-5.5-high",
+		reasoningEffort: "high"
+	}],
+	"claude-opus-5": [{
+		sourceModelId: "claude-opus-5-high",
+		reasoningEffort: "high"
+	}, {
+		sourceModelId: "claude-opus-5-max",
+		reasoningEffort: "max"
+	}],
+	"claude-opus-4-8": [{
+		sourceModelId: "claude-opus-4-8",
+		reasoningEffort: null
+	}, {
+		sourceModelId: "claude-opus-4-8-high",
+		reasoningEffort: "high"
+	}],
+	"claude-opus-4-7": [{
+		sourceModelId: "claude-opus-4-7",
+		reasoningEffort: null
+	}, {
+		sourceModelId: "claude-opus-4-7-high",
+		reasoningEffort: "high"
+	}],
+	"claude-sonnet-4-6": [{
+		sourceModelId: "claude-sonnet-4-6",
+		reasoningEffort: null
+	}],
+	"claude-opus-4-6": [{
+		sourceModelId: "claude-opus-4-6",
+		reasoningEffort: null
+	}, {
+		sourceModelId: "claude-opus-4-6-high",
+		reasoningEffort: "high"
+	}],
+	"claude-haiku-4-5-20251001": [{
+		sourceModelId: "claude-haiku-4-5-20251001",
+		reasoningEffort: null
+	}],
+	"us.anthropic.claude-sonnet-4-5-20250929-v1:0": [{
+		sourceModelId: "claude-sonnet-4-5-20250929",
+		reasoningEffort: null
+	}],
+	"us.anthropic.claude-haiku-4-5-20251001-v1:0": [{
+		sourceModelId: "claude-haiku-4-5-20251001",
+		reasoningEffort: null
+	}],
+	"us.amazon.nova-pro-v1:0": [{
+		sourceModelId: "amazon-nova-pro-v1.0",
+		reasoningEffort: null
+	}],
+	"us.amazon.nova-lite-v1:0": [{
+		sourceModelId: "amazon-nova-lite-v1.0",
+		reasoningEffort: null
+	}],
+	"us.amazon.nova-micro-v1:0": [{
+		sourceModelId: "amazon-nova-micro-v1.0",
+		reasoningEffort: null
+	}],
+	"openai.gpt-oss-120b-1:0": [{
+		sourceModelId: "gpt-oss-120b",
+		reasoningEffort: null
+	}],
+	"openai.gpt-oss-20b-1:0": [{
+		sourceModelId: "gpt-oss-20b",
+		reasoningEffort: null
+	}],
+	"mistral-large-4": [],
+	"mistral-large-latest": [],
+	"mistral-medium-latest": [],
+	"mistral-small-latest": []
+};
+const MODEL_UNRATED_REASONS = [
+	"too_new",
+	"preview_only",
+	"named_default_effort",
+	"floating_alias"
+];
+/**
+* A reviewed reason for every offered model without an exact Text Arena row.
+* A newly offered unrated model fails typecheck until it has one, and mapping
+* a source later makes its entry here an excess key, so the lists cannot
+* drift. `too_new` drives the picker's "New" recommendation.
+*/
+const MODEL_UNRATED_REASON = {
+	"gemini-3.1-flash-lite": "preview_only",
+	"openai/gpt-6.1-sol": "too_new",
+	"google/gemini-3.5-flash-lite": "named_default_effort",
+	"google/gemini-3.1-pro-preview": "named_default_effort",
+	"google/gemini-3.1-flash-lite": "preview_only",
+	"anthropic/claude-sonnet-5.5": "too_new",
+	"anthropic/claude-sonnet-4.6": "named_default_effort",
+	"gpt-6.1-sol": "too_new",
+	"claude-sonnet-5-5": "too_new",
+	"mistral-large-4": "too_new",
+	"mistral-large-latest": "floating_alias",
+	"mistral-medium-latest": "floating_alias",
+	"mistral-small-latest": "floating_alias"
+};
+//#endregion
+//#region src/benchmarks.gen.ts
+const MODEL_BENCHMARK_PUBLISH_DATE = "2026-09-30";
+const MODEL_BENCHMARK_RATINGS = {
+	"gemini-3.8-flash-high": {
+		rating: 1494.09,
+		ratingLower: 1489,
+		ratingUpper: 1499.18,
+		rank: 11,
+		voteCount: 24828
+	},
+	"gemini-3.7-flash-high": {
+		rating: 1488.3,
+		ratingLower: 1483.08,
+		ratingUpper: 1493.52,
+		rank: 16,
+		voteCount: 20851
+	},
+	"gemini-3.6-flash-high": {
+		rating: 1483.09,
+		ratingLower: 1478.64,
+		ratingUpper: 1487.53,
+		rank: 21,
+		voteCount: 35640
+	},
+	"gemini-3.5-flash-lite": {
+		rating: 1454.11,
+		ratingLower: 1449.62,
+		ratingUpper: 1458.6,
+		rank: 70,
+		voteCount: 35335
+	},
+	"gemini-3.1-pro-preview": {
+		rating: 1486.96,
+		ratingLower: 1483.95,
+		ratingUpper: 1489.98,
+		rank: 18,
+		voteCount: 121225
+	},
+	"gemini-3.5-flash-medium": {
+		rating: 1476.2,
+		ratingLower: 1472.04,
+		ratingUpper: 1480.36,
+		rank: 29,
+		voteCount: 46735
+	},
+	"gemini-3.5-flash-high": {
+		rating: 1477.07,
+		ratingLower: 1473.17,
+		ratingUpper: 1480.97,
+		rank: 27,
+		voteCount: 47981
+	},
+	"gpt-6-astra-max": {
+		rating: 1476.18,
+		ratingLower: 1469.17,
+		ratingUpper: 1483.19,
+		rank: 30,
+		voteCount: 8565
+	},
+	"gpt-6-sol-max": {
+		rating: 1455.82,
+		ratingLower: 1448.52,
+		ratingUpper: 1463.12,
+		rank: 66,
+		voteCount: 7574
+	},
+	"gpt-6-luna-max": {
+		rating: 1443.84,
+		ratingLower: 1436.67,
+		ratingUpper: 1451.02,
+		rank: 85,
+		voteCount: 7654
+	},
+	"gpt-5.6-sol-xhigh": {
+		rating: 1483.76,
+		ratingLower: 1479.3,
+		ratingUpper: 1488.22,
+		rank: 20,
+		voteCount: 36002
+	},
+	"gpt-5.6-terra-xhigh": {
+		rating: 1465.13,
+		ratingLower: 1460.7,
+		ratingUpper: 1469.56,
+		rank: 50,
+		voteCount: 37056
+	},
+	"gpt-5.6-luna-xhigh": {
+		rating: 1453.41,
+		ratingLower: 1448.99,
+		ratingUpper: 1457.82,
+		rank: 72,
+		voteCount: 37593
+	},
+	"claude-sonnet-5-high": {
+		rating: 1461.83,
+		ratingLower: 1457.61,
+		ratingUpper: 1466.05,
+		rank: 55,
+		voteCount: 45191
+	},
+	"claude-opus-5-high": {
+		rating: 1491.19,
+		ratingLower: 1487.38,
+		ratingUpper: 1495,
+		rank: 13,
+		voteCount: 58380
+	},
+	"claude-opus-5-max": {
+		rating: 1489.08,
+		ratingLower: 1484.09,
+		ratingUpper: 1494.07,
+		rank: 14,
+		voteCount: 28351
+	},
+	"claude-opus-4-8-high": {
+		rating: 1481.29,
+		ratingLower: 1477.4,
+		ratingUpper: 1485.17,
+		rank: 23,
+		voteCount: 63577
+	},
+	"gpt-5.5-high": {
+		rating: 1481.34,
+		ratingLower: 1477.61,
+		ratingUpper: 1485.07,
+		rank: 22,
+		voteCount: 69312
+	},
+	"gpt-5.4-mini-high": {
+		rating: 1447.05,
+		ratingLower: 1443.29,
+		ratingUpper: 1450.8,
+		rank: 81,
+		voteCount: 63311
+	},
+	"gpt-5.5": {
+		rating: 1476.73,
+		ratingLower: 1473.03,
+		ratingUpper: 1480.42,
+		rank: 28,
+		voteCount: 70894
+	},
+	"gpt-5.4": {
+		rating: 1464.86,
+		ratingLower: 1461.16,
+		ratingUpper: 1468.56,
+		rank: 51,
+		voteCount: 67753
+	},
+	"gpt-5.4-high": {
+		rating: 1475.18,
+		ratingLower: 1471.41,
+		ratingUpper: 1478.96,
+		rank: 33,
+		voteCount: 64648
+	},
+	"gpt-5.4-nano-high": {
+		rating: 1401.13,
+		ratingLower: 1397.34,
+		ratingUpper: 1404.92,
+		rank: 156,
+		voteCount: 62300
+	},
+	"gpt-5.2": {
+		rating: 1435.61,
+		ratingLower: 1432.43,
+		ratingUpper: 1438.79,
+		rank: 101,
+		voteCount: 83682
+	},
+	"gpt-5.2-high": {
+		rating: 1437.33,
+		ratingLower: 1433.77,
+		ratingUpper: 1440.89,
+		rank: 98,
+		voteCount: 49549
+	},
+	"claude-fable-5.1-max": {
+		rating: 1501.25,
+		ratingLower: 1494.75,
+		ratingUpper: 1507.75,
+		rank: 6,
+		voteCount: 11241
+	},
+	"claude-fable-5-high": {
+		rating: 1504.87,
+		ratingLower: 1500.46,
+		ratingUpper: 1509.28,
+		rank: 3,
+		voteCount: 37900
+	},
+	"claude-opus-5.5-high": {
+		rating: 1503.98,
+		ratingLower: 1494.32,
+		ratingUpper: 1513.64,
+		rank: 4,
+		voteCount: 3932
+	},
+	"claude-opus-4-8": {
+		rating: 1474.52,
+		ratingLower: 1470.67,
+		ratingUpper: 1478.38,
+		rank: 36,
+		voteCount: 64711
+	},
+	"claude-opus-4-7": {
+		rating: 1494.34,
+		ratingLower: 1490.57,
+		ratingUpper: 1498.11,
+		rank: 9,
+		voteCount: 65705
+	},
+	"claude-opus-4-7-high": {
+		rating: 1501.69,
+		ratingLower: 1497.92,
+		ratingUpper: 1505.46,
+		rank: 5,
+		voteCount: 64607
+	},
+	"claude-sonnet-4-6": {
+		rating: 1472.13,
+		ratingLower: 1468.61,
+		ratingUpper: 1475.65,
+		rank: 42,
+		voteCount: 70828
+	},
+	"claude-opus-4-6": {
+		rating: 1497.37,
+		ratingLower: 1494.03,
+		ratingUpper: 1500.7,
+		rank: 7,
+		voteCount: 81769
+	},
+	"claude-opus-4-6-high": {
+		rating: 1505.49,
+		ratingLower: 1502.06,
+		ratingUpper: 1508.91,
+		rank: 2,
+		voteCount: 77193
+	},
+	"claude-haiku-4-5-20251001": {
+		rating: 1413.56,
+		ratingLower: 1411.12,
+		ratingUpper: 1415.99,
+		rank: 139,
+		voteCount: 142743
+	},
+	"claude-sonnet-4-5-20250929": {
+		rating: 1455.18,
+		ratingLower: 1452.31,
+		ratingUpper: 1458.05,
+		rank: 68,
+		voteCount: 82438
+	},
+	"amazon-nova-pro-v1.0": {
+		rating: 1290.1,
+		ratingLower: 1285.51,
+		ratingUpper: 1294.69,
+		rank: 284,
+		voteCount: 24745
+	},
+	"amazon-nova-lite-v1.0": {
+		rating: 1260.43,
+		ratingLower: 1255.24,
+		ratingUpper: 1265.62,
+		rank: 312,
+		voteCount: 19372
+	},
+	"amazon-nova-micro-v1.0": {
+		rating: 1240.7,
+		ratingLower: 1235.52,
+		ratingUpper: 1245.88,
+		rank: 318,
+		voteCount: 19364
+	},
+	"gpt-oss-120b": {
+		rating: 1351.7,
+		ratingLower: 1347.32,
+		ratingUpper: 1356.08,
+		rank: 209,
+		voteCount: 30039
+	},
+	"gpt-oss-20b": {
+		rating: 1317.59,
+		ratingLower: 1311.16,
+		ratingUpper: 1324.01,
+		rank: 260,
+		voteCount: 10371
+	}
+};
+//#endregion
+//#region src/benchmarks.ts
+/**
+* Cost and quality trade-offs across the offered catalog.
+*
+* Quality is the Text Arena (LMArena) rating of an exactly matched model and
+* reasoning-effort variant (`benchmark-sources.ts` + `benchmarks.gen.ts`).
+* Cost is the list price of one typical call. The API classifies once over the
+* whole catalog (`benchmark-frontier.ts`) so clients never ship the rate table
+* or the snapshot.
+*/
+const TYPICAL_CALL_INPUT_TOKENS = 2e4;
+const TYPICAL_CALL_OUTPUT_TOKENS = 2e3;
+const TOKENS_PER_RATE_UNIT = 1e6;
+/**
+* List price in USD of one typical call at the standard (non-cached,
+* below-threshold) rate. Rates are per model, not per effort, so this does not
+* vary with reasoning effort even though higher efforts usually emit more
+* output tokens; the comparison states that assumption beside the chart.
+*/
+const getTypicalCallCostUsd = (modelId) => {
+	const rate = getModelRate(modelId);
+	if (rate === void 0) return null;
+	const { inputPerMTok, outputPerMTok } = getStandardModelRate(rate);
+	return (inputPerMTok * TYPICAL_CALL_INPUT_TOKENS + outputPerMTok * TYPICAL_CALL_OUTPUT_TOKENS) / TOKENS_PER_RATE_UNIT / MODEL_RATE_UNITS_PER_USD;
+};
+const MODEL_BENCHMARK_RATINGS_BY_ID = MODEL_BENCHMARK_RATINGS;
+/** Every measured variant of an offered model, in source order. */
+const getModelBenchmarkMeasurements = (modelId) => MODEL_BENCHMARK_SOURCES[modelId].flatMap(({ sourceModelId, reasoningEffort }) => {
+	const measured = MODEL_BENCHMARK_RATINGS_BY_ID[sourceModelId];
+	return measured === void 0 ? [] : [{
+		sourceModelId,
+		reasoningEffort,
+		rating: measured.rating,
+		ratingLower: measured.ratingLower,
+		ratingUpper: measured.ratingUpper
+	}];
+});
+const MODEL_UNRATED_REASON_BY_ID = MODEL_UNRATED_REASON;
+/** Why an offered model has no exact Text Arena row; `null` when it has one. */
+const getModelUnratedReason = (modelId) => MODEL_UNRATED_REASON_BY_ID[modelId] ?? null;
+//#endregion
+export { MODEL_BENCHMARK_CATEGORY, MODEL_BENCHMARK_CONFIG, MODEL_BENCHMARK_DATASET, MODEL_BENCHMARK_LICENCE, MODEL_BENCHMARK_NAME, MODEL_BENCHMARK_PUBLISH_DATE, MODEL_BENCHMARK_RATINGS, MODEL_BENCHMARK_SOURCES, MODEL_BENCHMARK_SOURCE_URL, MODEL_BENCHMARK_SPLIT, MODEL_UNRATED_REASON, MODEL_UNRATED_REASONS, TYPICAL_CALL_INPUT_TOKENS, TYPICAL_CALL_OUTPUT_TOKENS, getModelBenchmarkMeasurements, getModelUnratedReason, getTypicalCallCostUsd };