Skip to content

OpenAI's Codex Security CLI and TypeScript SDK for finding, validating, and fixing security vulnerabilities. npm: https://www.npmjs.com/package/@openai/codex-security

refactor(sdk): consolidate direct model pricing (#2470)

Michael D'Angelo57bf0539eadc
diff --git a/sdk/typescript/src/cost-model.ts b/sdk/typescript/src/cost-model.ts
index 705141bcb..b152d7955 100644
--- a/sdk/typescript/src/cost-model.ts
+++ b/sdk/typescript/src/cost-model.ts
@@ -53,42 +53,67 @@ export interface ScanTokenUsage {
total_tokens: number;
}
-const MODEL_PRICING_NANODOLLARS: Readonly<Record<string, ModelPricing>> = {
- // GPT-5.5 has no additional cache-write charge.
- "gpt-5.5": [5_000, 500, 5_000, 30_000],
- "gpt-5.5-2026-04-23": [5_000, 500, 5_000, 30_000],
- "gpt-6-sol": [2_000, 200, 2_500, 10_000],
- "gpt-6.1-sol": [2_000, 100, 2_500, 10_000],
- "gpt-6-astra": [10_000, 1_000, 12_500, 50_000],
- "gpt-6-luna": [100, 10, 125, 500],
- "gpt-5.6": [4_000, 400, 5_000, 20_000],
- "gpt-5.6-sol": [4_000, 400, 5_000, 20_000],
- "gpt-5.6-terra": [2_000, 200, 2_500, 12_000],
- "gpt-5.6-luna": [200, 20, 250, 1_200],
- // https://developers.openai.com/api/docs/pricing#cyber-models
- "gpt-5.6-cyber": [12_500, 1_250, 15_625, 75_000],
- "gpt-6-cyber": [20_000, 2_000, 25_000, 100_000],
-};
+// Direct Standard rates in nanodollars per token. Omit the long-context
+// rate when there is no verified upper estimate for a model.
+function directRate(
+ asOf: string,
+ short: ModelPricing,
+ long?: ModelPricing,
+ source = "https://developers.openai.com/api/docs/pricing",
+) {
+ return { short, long, unitsPerUsd: 1_000_000_000, source, asOf };
+}
-// Verified Standard rates: https://developers.openai.com/api/docs/pricing
-// GPT-5.5: https://developers.openai.com/api/docs/models/gpt-5.5
-// GPT-6 Sol: https://developers.openai.com/api/docs/models/gpt-6-sol
-// Do not infer tiers from aggregate scan tokens: the runtime does not report
-// which usage received long-context pricing. GPT-5.6-Cyber and GPT-6-Cyber have no
-// verified long-context rates, so their upper estimates stay unavailable.
-const LONG_CONTEXT_PRICING_NANODOLLARS: Readonly<Record<string, ModelPricing>> =
- {
- "gpt-5.5": [10_000, 1_000, 10_000, 45_000],
- "gpt-5.5-2026-04-23": [10_000, 1_000, 10_000, 45_000],
- "gpt-6-sol": [4_000, 400, 5_000, 15_000],
- "gpt-6.1-sol": [4_000, 200, 5_000, 15_000],
- "gpt-6-astra": [20_000, 2_000, 25_000, 75_000],
- "gpt-6-luna": [200, 20, 250, 750],
- "gpt-5.6": [8_000, 800, 10_000, 30_000],
- "gpt-5.6-sol": [8_000, 800, 10_000, 30_000],
- "gpt-5.6-terra": [4_000, 400, 5_000, 18_000],
- "gpt-5.6-luna": [400, 40, 500, 1_800],
- };
+const gpt55 = directRate(
+ "2026-09-14",
+ [5_000, 500, 5_000, 30_000],
+ [10_000, 1_000, 10_000, 45_000],
+ "https://developers.openai.com/api/docs/models/gpt-5.5",
+);
+const gpt56Sol = directRate(
+ "2026-09-14",
+ [4_000, 400, 5_000, 20_000],
+ [8_000, 800, 10_000, 30_000],
+);
+const directPricing: Readonly<Record<string, ReturnType<typeof directRate>>> = {
+ "gpt-5.5": gpt55,
+ "gpt-5.5-2026-04-23": gpt55,
+ "gpt-6-sol": directRate(
+ "2026-10-10",
+ [2_000, 200, 2_500, 10_000],
+ [4_000, 400, 5_000, 15_000],
+ "https://developers.openai.com/api/docs/models/gpt-6-sol",
+ ),
+ "gpt-6.1-sol": directRate(
+ "2026-09-30",
+ [2_000, 100, 2_500, 10_000],
+ [4_000, 200, 5_000, 15_000],
+ ),
+ "gpt-6-astra": directRate(
+ "2026-09-14",
+ [10_000, 1_000, 12_500, 50_000],
+ [20_000, 2_000, 25_000, 75_000],
+ ),
+ "gpt-6-luna": directRate(
+ "2026-09-30",
+ [100, 10, 125, 500],
+ [200, 20, 250, 750],
+ ),
+ "gpt-5.6": gpt56Sol,
+ "gpt-5.6-sol": gpt56Sol,
+ "gpt-5.6-terra": directRate(
+ "2026-09-14",
+ [2_000, 200, 2_500, 12_000],
+ [4_000, 400, 5_000, 18_000],
+ ),
+ "gpt-5.6-luna": directRate(
+ "2026-09-14",
+ [200, 20, 250, 1_200],
+ [400, 40, 500, 1_800],
+ ),
+ "gpt-5.6-cyber": directRate("2026-10-06", [12_500, 1_250, 15_625, 75_000]),
+ "gpt-6-cyber": directRate("2026-11-05", [20_000, 2_000, 25_000, 100_000]),
+};
function usdPerMillionTokens(pricing: ModelPricing, unitsPerUsd: number) {
const [input, cacheRead, cacheWrite, output] = pricing;
@@ -147,6 +172,8 @@ function providerModelPricing(
model: string,
context?: ScanPricingContext | string | null,
) {
+ if (context === undefined && Object.hasOwn(directPricing, model))
+ return directPricing[model];
if (context == null || typeof context === "string") {
return bedrockModelPricing(model, context);
}
@@ -170,16 +197,10 @@ export function estimateScanCost(
context?: ScanPricingContext | string | null,
): ScanCost | null {
if (model === undefined) return null;
- const providerPricing = providerModelPricing(model, context);
- const unitsPerUsd = providerPricing?.unitsPerUsd ?? 1_000_000_000;
- const pricing =
- providerPricing?.short ??
- (context === undefined && Object.hasOwn(MODEL_PRICING_NANODOLLARS, model)
- ? MODEL_PRICING_NANODOLLARS[model]
- : undefined);
+ const rate = providerModelPricing(model, context);
const normalized = tokenUsage(usage);
- if (pricing === undefined || normalized === null) return null;
- const [inputRate, cachedInputRate, cacheWriteInputRate, outputRate] = pricing;
+ if (rate === undefined || normalized === null) return null;
+ const { short, long, unitsPerUsd, source, asOf } = rate;
const {
input_tokens: inputTokens,
cached_input_tokens: cachedInputTokens,
@@ -187,75 +208,49 @@ export function estimateScanCost(
output_tokens: outputTokens,
} = normalized;
- const costUnits =
- (inputTokens - cachedInputTokens - cacheWriteInputTokens) * inputRate +
- cachedInputTokens * cachedInputRate +
- cacheWriteInputTokens * cacheWriteInputRate +
- outputTokens * outputRate;
- if (!Number.isSafeInteger(costUnits)) return null;
+ const uncachedTokens =
+ inputTokens - cachedInputTokens - cacheWriteInputTokens;
- const longPricing = providerPricing
- ? providerPricing.long
- : LONG_CONTEXT_PRICING_NANODOLLARS[model];
- let maximumUnits: number | null = null;
- if (longPricing !== undefined) {
- const [longInput, longRead, longWrite, longOutput] = longPricing;
- // Unclassified input may include additional cache writes. Preserve the
- // reported subtotal, but include that uncertainty in the upper estimate.
- const uncachedRate =
- normalized.cache_write_input_tokens_reported === false
- ? Math.max(longInput, longWrite)
- : longInput;
- const maximum =
- (inputTokens - cachedInputTokens - cacheWriteInputTokens) * uncachedRate +
- cachedInputTokens * longRead +
- cacheWriteInputTokens * longWrite +
- outputTokens * longOutput;
- if (Number.isSafeInteger(maximum)) maximumUnits = maximum;
+ function priceInUsd(
+ [input, read, write, output]: ModelPricing,
+ unknownWrites: boolean,
+ ) {
+ // Unclassified input may include writes. Account for them only in the upper bound.
+ const units =
+ uncachedTokens * (unknownWrites ? Math.max(input, write) : input) +
+ cachedInputTokens * read +
+ cacheWriteInputTokens * write +
+ outputTokens * output;
+ return Number.isSafeInteger(units) ? units / unitsPerUsd : null;
}
+ const estimatedUsd = priceInUsd(short, false);
+ if (estimatedUsd === null) return null;
+ const unknownWrites = normalized.cache_write_input_tokens_reported === false;
return {
model,
inputTokens,
cachedInputTokens,
cacheWriteInputTokens,
- ...(normalized.cache_write_input_tokens_reported === false
- ? { cacheWriteInputTokensReported: false }
- : {}),
+ ...(unknownWrites ? { cacheWriteInputTokensReported: false } : {}),
outputTokens,
- estimatedUsd: costUnits / unitsPerUsd,
+ estimatedUsd,
estimatedUsdRange: {
- min: costUnits / unitsPerUsd,
- max: maximumUnits === null ? null : maximumUnits / unitsPerUsd,
+ min: estimatedUsd,
+ max: long === undefined ? null : priceInUsd(long, unknownWrites),
context: "unknown",
},
pricing: {
- source: providerPricing
- ? providerPricing.source
- : model.startsWith("gpt-5.5")
- ? "https://developers.openai.com/api/docs/models/gpt-5.5"
- : model === "gpt-6-sol"
- ? "https://developers.openai.com/api/docs/models/gpt-6-sol"
- : "https://developers.openai.com/api/docs/pricing",
- asOf: providerPricing
- ? providerPricing.asOf
- : model === "gpt-6-sol"
- ? "2026-10-10"
- : model === "gpt-5.6-cyber"
- ? "2026-10-06"
- : model === "gpt-6-cyber"
- ? "2026-11-05"
- : model === "gpt-6.1-sol" || model === "gpt-6-luna"
- ? "2026-09-30"
- : "2026-09-14",
+ source,
+ asOf,
serviceTier: "standard",
context: "short",
- usdPerMillionTokens: usdPerMillionTokens(pricing, unitsPerUsd),
- ...(longPricing === undefined
+ usdPerMillionTokens: usdPerMillionTokens(short, unitsPerUsd),
+ ...(long === undefined
? {}
: {
longContextUsdPerMillionTokens: usdPerMillionTokens(
- longPricing,
+ long,
unitsPerUsd,
),
}),