Skip to content

🪢 Open source agent evals & observability: Trace, evaluate, and improve LLM applications with one open platform.

chore(pricing): add GPT-6.1 Sol pricing and GPT-6 Astra Ultrafast tier (#18055)

langfuse-bot646c15a60ebe
diff --git a/.agents/skills/add-model-price/SKILL.md b/.agents/skills/add-model-price/SKILL.md
index cf414bfeb7..87e131e3bf 100644
--- a/.agents/skills/add-model-price/SKILL.md
+++ b/.agents/skills/add-model-price/SKILL.md
@@ -34,6 +34,7 @@ updates in `packages/shared/`.
- Generate a lowercase UUID for the model entry.
- Create a `matchPattern` that covers supported provider formats.
- Add at least one default pricing tier.
+- Write every per-token price as `<USD per MTok>e-6` (`0.1e-6`, not `1e-7`).
- Map every supported semantic usage bucket to the provider aliases Langfuse
may persist.
- Insert the pricing entry into `worker/src/constants/default-model-prices.json`.
diff --git a/.agents/skills/add-model-price/references/automated-audit.md b/.agents/skills/add-model-price/references/automated-audit.md
index 55c6f99236..25bc36c5fd 100644
--- a/.agents/skills/add-model-price/references/automated-audit.md
+++ b/.agents/skills/add-model-price/references/automated-audit.md
@@ -67,6 +67,9 @@ report over an uncertain code change.
- Preserve existing IDs when updating an entry.
- Generate a new lowercase UUID for a new model entry.
- Refresh `updatedAt` only for entries that changed.
+- Write every per-token price in a changed entry as `<USD per MTok>e-6`
+ (`0.1e-6`, not `1e-7`; `25e-6`, not `2.5e-5`). See "Price Conversion" in
+ `provider-sources-and-price-keys.md`.
- Add newly released major models when they are officially documented, priced,
and representable with Langfuse's existing usage keys. Prioritize flagship
text/chat/reasoning models and models named by manual audit instructions.
diff --git a/.agents/skills/add-model-price/references/model-audit-memory.md b/.agents/skills/add-model-price/references/model-audit-memory.md
index b6c277fb19..b783fc3c12 100644
--- a/.agents/skills/add-model-price/references/model-audit-memory.md
+++ b/.agents/skills/add-model-price/references/model-audit-memory.md
@@ -12,11 +12,25 @@ audit date.
## Latest useful snapshot
-**Audit date:** 2026-09-29
+**Audit date:** 2026-09-30
All prices listed as `$X / MTok` (per million tokens). Per-token JSON values: divide by 1,000,000.
-The 2026-09-29 run re-fetched the full Anthropic pricing page (model table,
+The 2026-09-30 run re-fetched the full Anthropic pricing page (model table,
+cache-hits footnote, Fast mode and Batch tables), the Anthropic models-overview
+comparison table, the OpenAI aggregate Standard/Long-Context/Fast-mode/Flex
+pricing tables, the full OpenAI model catalog, both Gemini pricing pages (3.x
+and 2.5 families, with explicit Free/Paid column separation), the Gemini
+models catalog, the AWS Bedrock pricing page, and the TypeSafe Jev models
+page. This run found one new model: **GPT-6.1 Sol** (`gpt-6.1-sol`), a new
+sibling alongside the still-active `gpt-6-sol` — same $2/$10 input/output and
+$2.50 cache-write rate, but half the cached-input price ($0.10/MTok vs.
+$0.20/MTok). See `provider-sources-and-price-keys.md`'s "September 30 2026"
+entry for the full six-tier pricing and matchPattern detail. Every other price
+checked this run — including every existing Anthropic, OpenAI, and Gemini row
+below, the AWS Bedrock Public Extended Access limitation, and the TypeSafe Jev
+`jev-1.13.0` pricing — matched verbatim; no other updates were needed. The
+2026-09-29 run before it re-fetched the full Anthropic pricing page (model table,
cache-hits footnote, Fast mode and Batch tables), the Anthropic models-overview
comparison table, the OpenAI aggregate Standard/Fast-mode/Flex pricing summary,
the full OpenAI model catalog, a dedicated re-fetch of `gpt-5-chat-latest`'s own
@@ -72,8 +86,9 @@ to have siblings `gpt-5.5-cyber`/`gpt-5.4-cyber`, see provider-sources-and-price
| Provider | Model / pricing entry | Pricing checked | Price confirmed | Tiering checked | Tiering correct | Change | Official source(s) | Comments |
| -------- | --------------------- | --------------- | --------------- | --------------- | --------------- | ------ | ------------------- | -------- |
-| OpenAI | gpt-6-astra | Input $10/MTok, Cached $1/MTok, Cache write $12.50/MTok, Output $50/MTok | Yes | Large Context (>272K) 2x/2x/2x/1.5x; Fast mode 2x base; Flex 0.5x base | Yes | None | https://developers.openai.com/api/docs/pricing https://developers.openai.com/api/docs/models/gpt-6-astra | Added 2026-09-03, re-confirmed unchanged 2026-09-10, 2026-09-14, and 2026-09-22. Six-tier gpt-5.6-sol-style key set. |
-| OpenAI | gpt-6-sol | Input $2/MTok, Cached $0.20/MTok, Cache write $2.50/MTok, Output $10/MTok | Yes | Large Context (>272K) 2x/2x/2x/1.5x; Fast mode 2x applicable tier; Flex 0.5x applicable tier | Yes | Added | https://developers.openai.com/api/docs/pricing https://developers.openai.com/api/docs/models/gpt-6-sol | Added 2026-09-22. Six-tier gpt-6-astra-style key set; same 1,050,000-token context window and 272K large-context threshold as gpt-6-astra. |
+| OpenAI | gpt-6-astra | Input $10/MTok, Cached $1/MTok, Cache write $12.50/MTok, Output $50/MTok | Yes | Large Context (>272K) 2x/2x/2x/1.5x; Fast mode 2x base; Flex 0.5x base; Ultrafast 6x base ($60/$6/$75/$300, long-context $120/$12/$150/$450) | Yes | Updated | https://developers.openai.com/api/docs/pricing https://developers.openai.com/api/docs/models/gpt-6-astra https://developers.openai.com/api/docs/guides/ultrafast-mode | Added 2026-09-03, re-confirmed unchanged 2026-09-10, 2026-09-14, and 2026-09-22. Ultrafast (`service_tier: "ultrafast"`) and Ultrafast · Large context tiers added 2026-09-30; it is the only model on the Ultrafast pricing tab. Eight-tier gpt-5.6-sol-style key set. |
+| OpenAI | gpt-6-sol | Input $2/MTok, Cached $0.20/MTok, Cache write $2.50/MTok, Output $10/MTok | Yes | Large Context (>272K) 2x/2x/2x/1.5x; Fast mode 2x applicable tier; Flex 0.5x applicable tier | Yes | None | https://developers.openai.com/api/docs/pricing https://developers.openai.com/api/docs/models/gpt-6-sol | Added 2026-09-22, re-confirmed unchanged 2026-09-30. Six-tier gpt-6-astra-style key set; same 1,050,000-token context window and 272K large-context threshold as gpt-6-astra. Still a distinct, separately-priced model ID from the newer gpt-6.1-sol. |
+| OpenAI | gpt-6.1-sol | Input $2/MTok, Cached $0.10/MTok, Cache write $2.50/MTok, Output $10/MTok | Yes | Large Context (>272K) 2x/2x/2x/1.5x; Fast mode 2x applicable tier (large-context Fast row directly quoted); Flex 0.5x applicable tier (Flex·Large-Context derived, not separately published) | Yes | Added | https://developers.openai.com/api/docs/pricing https://developers.openai.com/api/docs/models/gpt-6.1-sol | Added 2026-09-30. Six-tier gpt-6-sol-style key set; same context window and threshold as gpt-6-sol/gpt-6-astra. Cached-input price is half of gpt-6-sol's ($0.10 vs $0.20/MTok); every other base rate is identical. |
| OpenAI | gpt-6-luna | Input $0.10/MTok, Cached $0.01/MTok, Cache write $0.125/MTok, Output $0.50/MTok | Yes | Large Context (>272K) 2x/2x/2x/1.5x; Fast mode 2x applicable tier; Flex 0.5x applicable tier | Yes | Added | https://developers.openai.com/api/docs/pricing https://developers.openai.com/api/docs/models/gpt-6-luna | Added 2026-09-22. Six-tier gpt-6-astra-style key set; same context window and threshold as gpt-6-astra/gpt-6-sol. |
| Provider | Model / pricing entry | Pricing checked | Price confirmed | Tiering checked | Tiering correct | Change | Official source(s) | Comments |
@@ -98,7 +113,7 @@ to have siblings `gpt-5.5-cyber`/`gpt-5.4-cyber`, see provider-sources-and-price
| Anthropic | claude-haiku-4-5-20251001 | Input $1/MTok, Output $5/MTok, 5m $1.25/MTok, 1h $2/MTok, read $0.10/MTok | Yes | No large-context tier | Not applicable | None | https://platform.claude.com/docs/en/about-claude/pricing | Re-confirmed unchanged. |
| Anthropic | claude-3-5-haiku-20241022 | Input $0.80/MTok, Output $4/MTok, 5m $1/MTok, 1h $1.60/MTok, read $0.08/MTok | Yes | Retired except Bedrock/Google Cloud | Not applicable | None | https://platform.claude.com/docs/en/about-claude/pricing | Re-confirmed present, unchanged. |
| Anthropic | claude-3.7-sonnet-20250219 / claude-3.5-sonnet-20241022 / claude-3-5-sonnet-20240620 / claude-3-opus-20240229 / claude-3-sonnet-20240229 / claude-3-haiku-20240307 | Legacy prices retained | No | Not on current page | Not applicable | None | https://platform.claude.com/docs/en/about-claude/pricing | Not re-verified this run; retired, out of "flagship" scope. |
-| AWS Bedrock | claude-3-5-sonnet-20240620 / claude-3.5-sonnet-20241022 (Public Extended Access SKU) | $6.00/MTok input, $30.00/MTok output, $7.50/MTok cache write, $0.60/MTok cache read | Yes (confirmed real Aug 4 2026) | Distinct dated SKU, not a context-length tier | Not applicable | Unresolved | https://aws.amazon.com/bedrock/pricing/ | Not re-verified this run; permanent documented limitation (model-ID string match cannot distinguish billing SKU). |
+| AWS Bedrock | claude-3-5-sonnet-20240620 / claude-3.5-sonnet-20241022 (Public Extended Access SKU) | $6.00/MTok input, $30.00/MTok output, $7.50/MTok cache write, $0.60/MTok cache read | Yes (re-confirmed 2026-09-30) | Distinct dated SKU, not a context-length tier | Not applicable | Unresolved | https://aws.amazon.com/bedrock/pricing/ | Re-confirmed unchanged 2026-09-30; permanent documented limitation (model-ID string match cannot distinguish billing SKU). |
| OpenAI | gpt-5.6-sol | Input $4/MTok, Cached $0.40/MTok, Cache write $5.00/MTok, Output $20/MTok | Yes | Large Context (>272K): $8/$0.80/$10.00/$30; Fast mode 2x base; Flex 0.5x base | Yes | None | https://developers.openai.com/api/docs/pricing https://developers.openai.com/api/docs/models/gpt-5.6-sol | Promotional price cut from Aug 24 2026 reconfirmed still active, still "available at least through November 21, 2026" — re-verify after that date. |
| OpenAI | gpt-5.6-terra | Input $2/MTok, Cached $0.20/MTok, Cache write $2.50/MTok, Output $12/MTok | Yes | Large Context (>272K): $4/$0.40/$5.00/$18 | Yes | None | https://developers.openai.com/api/docs/pricing | Re-confirmed unchanged. |
| OpenAI | gpt-5.6-luna | Input $0.20/MTok, Cached $0.02/MTok, Cache write $0.25/MTok, Output $1.20/MTok | Yes | Large Context (>272K): $0.40/$0.04/$0.50/$1.80 | Yes | None | https://developers.openai.com/api/docs/pricing | Re-confirmed unchanged. |
diff --git a/.agents/skills/add-model-price/references/provider-sources-and-price-keys.md b/.agents/skills/add-model-price/references/provider-sources-and-price-keys.md
index 2626766cd9..40f5231896 100644
--- a/.agents/skills/add-model-price/references/provider-sources-and-price-keys.md
+++ b/.agents/skills/add-model-price/references/provider-sources-and-price-keys.md
@@ -394,6 +394,16 @@ file and `openAIModels`in July 27 2026 audit. Official sources:`https://develope
not individually re-quoted during the audit; all were re-read from the live table
before implementation). The pricing file now represents these with
`modelParameters.service_tier in ["flex"]`.
+ - **Ultrafast** (`service_tier: "ultrafast"`, documented September 30 2026) —
+ OpenAI's fastest tier, guide at
+ `https://developers.openai.com/api/docs/guides/ultrafast-mode`. The pricing
+ page's "Ultrafast" tab lists only `gpt-6-astra`: $60/$6/$75/$300 short-context
+ and $120/$12/$150/$450 long-context (>272K) per MTok input / cached input /
+ cache writes / output, i.e. 6x Standard. `gpt-6-astra` carries an `Ultrafast`
+ tier and an `Ultrafast · Large context (>272K)` tier, ordered with the other
+ combined service-tier + large-context tiers ahead of the single-condition
+ tiers. `gpt-5.6-sol` has preview-only Ultrafast access with no published
+ price; do not add a tier for it until the pricing page lists one.
- **Anthropic has the same class of gap**: the pricing page's "Fast mode pricing"
section documents Claude Opus 5 / Claude Opus 4.8 Fast mode at $10/$50 per MTok
input/output (`speed: "fast"` request parameter), but neither `claude-opus-5` nor
@@ -812,6 +822,40 @@ file and `openAIModels`in July 27 2026 audit. Official sources:`https://develope
Bedrock Public Extended Access SKU, and the legacy Claude 3.x/Gemini 1.x
catalog tail were not re-checked this run — no new evidence, standing
exclusions.
+- **GPT-6.1 Sol (added September 30 2026)** — OpenAI released `gpt-6.1-sol` as
+ a distinct sibling alongside the existing `gpt-6-sol` (both remain active,
+ separately priced model IDs — the model's own dedicated page directs users
+ from the old page to the new one, but the old `gpt-6-sol` page and pricing
+ row are still live). Confirmed via
+ `https://developers.openai.com/api/docs/pricing` (Standard, Long-Context,
+ Fast mode, and Flex tables, each independently re-quoted for this specific
+ model ID) and `https://developers.openai.com/api/docs/models/gpt-6.1-sol`
+ ("delivers near-Astra performance at a lower cost for complex coding,
+ computer use, and professional work"). Same 1,050,000-token context window
+ (max input 922,000, max output 128,000) and the same >272,000-input-token
+ Large Context threshold as the rest of the GPT-6 family. Pricing is
+ identical to `gpt-6-sol` for input ($2/MTok), cache write ($2.50/MTok), and
+ output ($10/MTok) — but **cached input is $0.10/MTok, half of `gpt-6-sol`'s
+ $0.20/MTok** (confirmed via three independent quotes of the Standard table
+ row: `gpt-6.1-sol | $2.00 | $0.10 | $2.50 | $10.00`). Large Context
+ (2x input/cache, 1.5x output): $4/$0.20/$5.00/$15. Fast mode
+ (`service_tier` in `["fast","priority"]`, 2x applicable tier): $4/$0.20/$5.00/$20
+ standard, $8/$0.40/$10.00/$30 large context (the large-context Fast row was
+ directly quoted from the aggregate table, not just derived from the
+ multiplier). Flex (`service_tier: "flex"`, 0.5x applicable tier):
+ $1.00/$0.05/$1.25/$5.00 standard; no separate Flex-Large-Context row is
+ published for this model (same gap as every other GPT-6-family member), so
+ the Flex·Large-Context tier ($2/$0.10/$2.50/$7.50) was derived by applying
+ the documented 0.5x multiplier to the Large Context tier, following the
+ `gpt-6-sol`/`gpt-6-astra` precedent. No date-stamped snapshot at launch.
+ Added to the pricing file mirroring `gpt-6-sol`'s exact six-tier key set and
+ to `openAIModels` in `types.ts` immediately after `gpt-6-sol` (not as the
+ first entry). matchPattern: `(?i)^(openai/)?(gpt-6.1-sol)$` — verified via
+ the bundled match-pattern tester that it does not collide with `gpt-6-sol`,
+ `gpt-6-luna`, or `gpt-6-astra` (all fully anchored with `^...$`; the
+ unescaped `.` follows this file's existing convention for other dotted
+ version numbers such as `gpt-5.6-sol` and `gemini-3.7-flash`, which likewise
+ do not escape the literal dot).
- **September 29 2026 audit: Claude Sonnet 5.5 added, released the day before
this run; everything else confirmed unchanged** — Re-fetched the full
Anthropic pricing page (model table, cache-hits footnote, Fast mode and
@@ -847,17 +891,29 @@ Capture:
Values in `default-model-prices.json` are per token, not per million tokens.
-| Provider Price | JSON Value |
-| -------------- | ---------- |
-| `$5 / MTok` | `5e-6` |
-| `$25 / MTok` | `25e-6` |
-| `$0.50 / MTok` | `0.5e-6` |
-| `$6.25 / MTok` | `6.25e-6` |
+Always write per-token prices as `<USD per MTok>e-6`, so the mantissa reads
+directly as the provider's per-million-token price. Keep the `e-6` exponent even
+for sub-dollar and sub-cent prices; never normalize to `1e-7`, `5e-8`, `1.5e-5`,
+or plain decimals like `0.000001`. `0` is fine for free usage types. The only
+exception is non-token usage (Gemini `grounding_queries` / `web_search_queries`
+per-query prices, written as `14e-3` for $14 per 1K queries). The validator
+enforces this on changed and selected entries; older entries may still use other
+notation and are rewritten only when that entry changes for another reason.
+
+| Provider Price | JSON Value | Not |
+| --------------- | ---------- | ----------- |
+| `$5 / MTok` | `5e-6` | |
+| `$25 / MTok` | `25e-6` | `2.5e-5` |
+| `$0.50 / MTok` | `0.5e-6` | `5e-7` |
+| `$0.10 / MTok` | `0.1e-6` | `1e-7` |
+| `$0.05 / MTok` | `0.05e-6` | `5e-8` |
+| `$6.25 / MTok` | `6.25e-6` | `0.00000625` |
Formula:
```text
price_per_token = price_per_mtok / 1_000_000
+json_literal = "<price_per_mtok>e-6"
```
- **TypeSafe Jev pricing (documented September 24 2026)** — `jev` is
diff --git a/.agents/skills/add-model-price/scripts/validate-pricing-file.mjs b/.agents/skills/add-model-price/scripts/validate-pricing-file.mjs
index 3d2d57bc89..b0c5812d88 100644
--- a/.agents/skills/add-model-price/scripts/validate-pricing-file.mjs
+++ b/.agents/skills/add-model-price/scripts/validate-pricing-file.mjs
@@ -52,6 +52,44 @@ function compileMatchPattern(rawPattern, label) {
}
}
+// Per-query prices (e.g. $14 per 1K grounding queries) are not per-token and
+// keep their natural notation.
+const NON_TOKEN_PRICE_KEYS = new Set([
+ "grounding_queries",
+ "groundingQueries",
+ "web_search_queries",
+ "webSearchQueries",
+]);
+const TOKEN_PRICE_LITERAL = /^(0|[0-9]+(\.[0-9]+)?e-6)$/;
+
+const MODEL_NAME_KEY = /"modelName"\s*:\s*("(?:[^"\\]|\\.)*")/g;
+
+// JSON.parse drops number notation, so this reads the raw source text.
+function validatePriceNotation(rawText, modelName) {
+ const modelNameMatches = [...rawText.matchAll(MODEL_NAME_KEY)];
+ const index = modelNameMatches.findIndex(
+ (match) => JSON.parse(match[1]) === modelName,
+ );
+ if (index === -1) return;
+ const block = rawText.slice(
+ modelNameMatches[index].index,
+ modelNameMatches[index + 1]?.index ?? rawText.length,
+ );
+
+ for (const pricesMatch of block.matchAll(/"prices"\s*:\s*\{([^}]*)\}/g)) {
+ for (const [, key, literal] of pricesMatch[1].matchAll(
+ /"([^"]+)"\s*:\s*([^,\s}]+)/g,
+ )) {
+ if (NON_TOKEN_PRICE_KEYS.has(key)) continue;
+ if (!TOKEN_PRICE_LITERAL.test(literal)) {
+ failures.push(
+ `${modelName}: price for ${key} must be written as <USD per 1M tokens>e-6 (got ${literal})`,
+ );
+ }
+ }
+ }
+}
+
function keysOfPrices(prices) {
return Object.keys(prices).sort();
}
@@ -415,6 +453,7 @@ for (const model of models) {
if (usageKeyModels.has(model.modelName)) {
validateUsageKeyCoverage(model);
+ validatePriceNotation(raw, model.modelName);
}
}
diff --git a/.agents/skills/add-model-price/scripts/validate-pricing-file.test.mjs b/.agents/skills/add-model-price/scripts/validate-pricing-file.test.mjs
index 7c96cb1c7e..7bd3ef7880 100644
--- a/.agents/skills/add-model-price/scripts/validate-pricing-file.test.mjs
+++ b/.agents/skills/add-model-price/scripts/validate-pricing-file.test.mjs
@@ -29,7 +29,23 @@ function model({ id, modelName, matchPattern, tokenizerId = null, prices }) {
};
}
-async function runValidator(currentModels, baseModels) {
+// Writes prices the way the pricing file stores them (`5e-6`), which
+// JSON.stringify alone would render as `0.000005`.
+function toPricingJson(models) {
+ const json = JSON.stringify(models, (key, value) =>
+ key === "prices"
+ ? Object.fromEntries(
+ Object.entries(value).map(([usageType, price]) => [
+ usageType,
+ `__per_mtok__${Number((price * 1e6).toFixed(6))}`,
+ ]),
+ )
+ : value,
+ );
+ return json.replace(/"__per_mtok__([0-9.]+)"/g, "$1e-6");
+}
+
+async function runValidator(currentModels, baseModels, { rawCurrent } = {}) {
const temporaryDirectory = await mkdtemp(
path.join(os.tmpdir(), "pricing-validator-"),
);
@@ -38,8 +54,8 @@ async function runValidator(currentModels, baseModels) {
try {
await Promise.all([
- writeFile(currentPath, JSON.stringify(currentModels)),
- writeFile(basePath, JSON.stringify(baseModels)),
+ writeFile(currentPath, rawCurrent ?? toPricingJson(currentModels)),
+ writeFile(basePath, toPricingJson(baseModels)),
]);
return spawnSync(
process.execPath,
@@ -190,6 +206,53 @@ test("rejects unequal prices within a semantic alias family", async () => {
assert.match(result.stderr, /cache-read aliases must have the same price/);
});
+test("requires per-million-token e-6 notation on changed entries", async () => {
+ const entry = model({
+ id: "bedrock-notation",
+ modelName: "amazon-nova-notation",
+ matchPattern: "(?i)^amazon-nova-notation$",
+ prices: { input: 1e-6, output: 5e-6 },
+ });
+ const withCachedInput = (literal) =>
+ toPricingJson([entry]).replace(
+ '"output":5e-6',
+ `"output":5e-6,"input_cached_tokens":${literal}`,
+ );
+
+ const offScale = await runValidator([], [], {
+ rawCurrent: withCachedInput("1e-7"),
+ });
+ assert.equal(offScale.status, 1);
+ assert.match(
+ offScale.stderr,
+ /amazon-nova-notation: price for input_cached_tokens must be written as <USD per 1M tokens>e-6 \(got 1e-7\)/,
+ );
+
+ const plainDecimal = await runValidator([], [], {
+ rawCurrent: withCachedInput("0.0000001"),
+ });
+ assert.equal(plainDecimal.status, 1);
+
+ const perMillion = await runValidator([], [], {
+ rawCurrent: withCachedInput("0.1e-6"),
+ });
+ assert.equal(perMillion.status, 0, perMillion.stderr);
+
+ const legacyEntry = {
+ ...entry,
+ pricingTiers: [
+ {
+ ...entry.pricingTiers[0],
+ prices: { input: 1e-6, output: 5e-6, input_cached_tokens: 1e-7 },
+ },
+ ],
+ };
+ const unchanged = await runValidator([], [legacyEntry], {
+ rawCurrent: withCachedInput("1e-7"),
+ });
+ assert.equal(unchanged.status, 0, unchanged.stderr);
+});
+
test("rejects an unknown explicitly selected model", () => {
const result = spawnSync(
process.execPath,
diff --git a/packages/shared/src/server/llm/types.ts b/packages/shared/src/server/llm/types.ts
index 01b22603dd..29aeffdb24 100644
--- a/packages/shared/src/server/llm/types.ts
+++ b/packages/shared/src/server/llm/types.ts
@@ -331,6 +331,7 @@ export const openAIModels = [
"gpt-4.1-nano-2025-04-14",
"gpt-6-astra",
"gpt-6-sol",
+ "gpt-6.1-sol",
"gpt-6-luna",
"gpt-5.6-sol",
"gpt-5.6-terra",
diff --git a/worker/src/constants/default-model-prices.json b/worker/src/constants/default-model-prices.json
index 8b88ba4e62..500431b846 100644
--- a/worker/src/constants/default-model-prices.json
+++ b/worker/src/constants/default-model-prices.json
@@ -7065,7 +7065,7 @@
"modelName": "gpt-6-astra",
"matchPattern": "(?i)^(openai/)?(gpt-6-astra)$",
"createdAt": "2026-09-03T00:00:00.000Z",
- "updatedAt": "2026-09-25T00:00:00.000Z",
+ "updatedAt": "2026-09-30T00:00:00.000Z",
"tokenizerConfig": {
"tokensPerName": 1,
"tokenizerModel": "gpt-4",
@@ -7093,11 +7093,44 @@
"reasoning_tokens": 50e-6
}
},
+ {
+ "id": "80d685ba-43fd-4f6f-94a9-c0558c0c8c1e",
+ "name": "Ultrafast · Large context (>272K)",
+ "isDefault": false,
+ "priority": 1,
+ "conditions": [
+ {
+ "source": "model_parameters",
+ "key": "service_tier",
+ "operator": "in",
+ "values": ["ultrafast"]
+ },
+ {
+ "usageDetailPattern": "(input|cache_write)",
+ "operator": "gt",
+ "value": 272000,
+ "caseSensitive": false
+ }
+ ],
+ "prices": {
+ "input": 120e-6,
+ "input_cached_tokens": 12e-6,
+ "input_cache_read": 12e-6,
+ "cache_read_input_tokens": 12e-6,
+ "input_cache_creation": 150e-6,
+ "cache_write_tokens": 150e-6,
+ "input_cache_write_tokens": 150e-6,
+ "output": 450e-6,
+ "output_reasoning_tokens": 450e-6,
+ "output_reasoning": 450e-6,
+ "reasoning_tokens": 450e-6
+ }
+ },
{
"id": "d3151184-4443-430a-a4ed-961dd55d223c",
"name": "Fast mode · Large context (>272K)",
"isDefault": false,
- "priority": 1,
+ "priority": 2,
"conditions": [
{
"source": "model_parameters",
@@ -7130,7 +7163,7 @@
"id": "1b8d0945-bd74-4223-8bed-320f2063a2db",
"name": "Flex · Large context (>272K)",
"isDefault": false,
- "priority": 2,
+ "priority": 3,
"conditions": [
{
"source": "model_parameters",
@@ -7159,11 +7192,38 @@
"reasoning_tokens": 37.5e-6
}
},
+ {
+ "id": "ad3d6431-99df-4090-b380-936a39502520",
+ "name": "Ultrafast",
+ "isDefault": false,
+ "priority": 4,
+ "conditions": [
+ {
+ "source": "model_parameters",
+ "key": "service_tier",
+ "operator": "in",
+ "values": ["ultrafast"]
+ }
+ ],
+ "prices": {
+ "input": 60e-6,
+ "input_cached_tokens": 6e-6,
+ "input_cache_read": 6e-6,
+ "cache_read_input_tokens": 6e-6,
+ "input_cache_creation": 75e-6,
+ "cache_write_tokens": 75e-6,
+ "input_cache_write_tokens": 75e-6,
+ "output": 300e-6,
+ "output_reasoning_tokens": 300e-6,
+ "output_reasoning": 300e-6,
+ "reasoning_tokens": 300e-6
+ }
+ },
{
"id": "2da4af90-9c52-452a-b9d2-73e8306bad6e",
"name": "Fast mode",
"isDefault": false,
- "priority": 3,
+ "priority": 5,
"conditions": [
{
"source": "model_parameters",
@@ -7190,7 +7250,7 @@
"id": "ec8b3a8e-86c2-4a31-8c60-5f8c4ade98a8",
"name": "Flex",
"isDefault": false,
- "priority": 4,
+ "priority": 6,
"conditions": [
{
"source": "model_parameters",
@@ -7217,7 +7277,7 @@
"id": "296fbd59-2e9a-43c0-8308-cbc3b9014fdd",
"name": "Large Context (>272K)",
"isDefault": false,
- "priority": 5,
+ "priority": 7,
"conditions": [
{
"usageDetailPattern": "(input|cache_write)",
@@ -7424,6 +7484,198 @@
}
]
},
+ {
+ "id": "ee39f703-e13e-43bc-b366-78596786cd06",
+ "modelName": "gpt-6.1-sol",
+ "matchPattern": "(?i)^(openai/)?(gpt-6.1-sol)$",
+ "createdAt": "2026-09-30T00:00:00.000Z",
+ "updatedAt": "2026-09-30T00:00:00.000Z",
+ "tokenizerConfig": {
+ "tokensPerName": 1,
+ "tokenizerModel": "gpt-4",
+ "tokensPerMessage": 3
+ },
+ "tokenizerId": "openai",
+ "pricingTiers": [
+ {
+ "id": "ee39f703-e13e-43bc-b366-78596786cd06_tier_default",
+ "name": "Standard",
+ "isDefault": true,
+ "priority": 0,
+ "conditions": [],
+ "prices": {
+ "input": 2e-6,
+ "input_cached_tokens": 0.1e-6,
+ "input_cache_read": 0.1e-6,
+ "cache_read_input_tokens": 0.1e-6,
+ "input_cache_creation": 2.5e-6,
+ "cache_write_tokens": 2.5e-6,
+ "input_cache_write_tokens": 2.5e-6,
+ "output": 10e-6,
+ "output_reasoning_tokens": 10e-6,
+ "output_reasoning": 10e-6,
+ "reasoning_tokens": 10e-6
+ }
+ },
+ {
+ "id": "b1a4b800-237b-4439-8f33-ad4457fcddff",
+ "name": "Fast mode · Large context (>272K)",
+ "isDefault": false,
+ "priority": 1,
+ "conditions": [
+ {
+ "source": "model_parameters",
+ "key": "service_tier",
+ "operator": "in",
+ "values": [
+ "fast",
+ "priority"
+ ]
+ },
+ {
+ "usageDetailPattern": "(input|cache_write)",
+ "operator": "gt",
+ "value": 272000,
+ "caseSensitive": false
+ }
+ ],
+ "prices": {
+ "input": 8e-6,
+ "input_cached_tokens": 0.4e-6,
+ "input_cache_read": 0.4e-6,
+ "cache_read_input_tokens": 0.4e-6,
+ "input_cache_creation": 10e-6,
+ "cache_write_tokens": 10e-6,
+ "input_cache_write_tokens": 10e-6,
+ "output": 30e-6,
+ "output_reasoning_tokens": 30e-6,
+ "output_reasoning": 30e-6,
+ "reasoning_tokens": 30e-6
+ }
+ },
+ {
+ "id": "80b5b730-df56-49e7-9c7d-efef4b312b53",
+ "name": "Flex · Large context (>272K)",
+ "isDefault": false,
+ "priority": 2,
+ "conditions": [
+ {
+ "source": "model_parameters",
+ "key": "service_tier",
+ "operator": "in",
+ "values": [
+ "flex"
+ ]
+ },
+ {
+ "usageDetailPattern": "(input|cache_write)",
+ "operator": "gt",
+ "value": 272000,
+ "caseSensitive": false
+ }
+ ],
+ "prices": {
+ "input": 2e-6,
+ "input_cached_tokens": 0.1e-6,
+ "input_cache_read": 0.1e-6,
+ "cache_read_input_tokens": 0.1e-6,
+ "input_cache_creation": 2.5e-6,
+ "cache_write_tokens": 2.5e-6,
+ "input_cache_write_tokens": 2.5e-6,
+ "output": 7.5e-6,
+ "output_reasoning_tokens": 7.5e-6,
+ "output_reasoning": 7.5e-6,
+ "reasoning_tokens": 7.5e-6
+ }
+ },
+ {
+ "id": "ff06c056-2f10-4d40-aa8b-96d7bffc0306",
+ "name": "Fast mode",
+ "isDefault": false,
+ "priority": 3,
+ "conditions": [
+ {
+ "source": "model_parameters",
+ "key": "service_tier",
+ "operator": "in",
+ "values": [
+ "fast",
+ "priority"
+ ]
+ }
+ ],
+ "prices": {
+ "input": 4e-6,
+ "input_cached_tokens": 0.2e-6,
+ "input_cache_read": 0.2e-6,
+ "cache_read_input_tokens": 0.2e-6,
+ "input_cache_creation": 5e-6,
+ "cache_write_tokens": 5e-6,
+ "input_cache_write_tokens": 5e-6,
+ "output": 20e-6,
+ "output_reasoning_tokens": 20e-6,
+ "output_reasoning": 20e-6,
+ "reasoning_tokens": 20e-6
+ }
+ },
+ {
+ "id": "1591fa11-7e04-4463-8998-cf4df6ded3af",
+ "name": "Flex",
+ "isDefault": false,
+ "priority": 4,
+ "conditions": [
+ {
+ "source": "model_parameters",
+ "key": "service_tier",
+ "operator": "in",
+ "values": [
+ "flex"
+ ]
+ }
+ ],
+ "prices": {
+ "input": 1e-6,
+ "input_cached_tokens": 0.05e-6,
+ "input_cache_read": 0.05e-6,
+ "cache_read_input_tokens": 0.05e-6,
+ "input_cache_creation": 1.25e-6,
+ "cache_write_tokens": 1.25e-6,
+ "input_cache_write_tokens": 1.25e-6,
+ "output": 5e-6,
+ "output_reasoning_tokens": 5e-6,
+ "output_reasoning": 5e-6,
+ "reasoning_tokens": 5e-6
+ }
+ },
+ {
+ "id": "c3578599-53f0-47be-8e79-802f3d81df23",
+ "name": "Large Context (>272K)",
+ "isDefault": false,
+ "priority": 5,
+ "conditions": [
+ {
+ "usageDetailPattern": "(input|cache_write)",
+ "operator": "gt",
+ "value": 272000,
+ "caseSensitive": false
+ }
+ ],
+ "prices": {
+ "input": 4e-6,
+ "input_cached_tokens": 0.2e-6,
+ "input_cache_read": 0.2e-6,
+ "cache_read_input_tokens": 0.2e-6,
+ "input_cache_creation": 5e-6,
+ "cache_write_tokens": 5e-6,
+ "input_cache_write_tokens": 5e-6,
+ "output": 15e-6,
+ "output_reasoning_tokens": 15e-6,
+ "output_reasoning": 15e-6,
+ "reasoning_tokens": 15e-6
+ }
+ }
+ ]
+ },
{
"id": "8edc337e-ba50-4811-bc51-e9ad4253826c",
"modelName": "gpt-6-luna",