Merge remote-tracking branch 'upstream/dev' into dev

Author: github-actions[bot]

packages/core/script/generate-friendli.ts

@@ -48,7 +48,6 @@ const familyPatterns: [RegExp, string][] = [
   [/llama-4/i, "llama-4"],
   [/qwen3/i, "qwen3"],
   [/deepseek-r1/i, "deepseek-r1"],
-  [/exaone/i, "exaone"],
   [/glm-4/i, "glm-4"],
   [/glm-5/i, "glm"],
 ];

packages/core/script/generate-venice.ts

@@ -3,30 +3,11 @@
 import { z } from "zod";
 import path from "node:path";
 import { readdir } from "node:fs/promises";
-import * as readline from "node:readline";
 import { ModelFamilyValues } from "../src/family.js";
 
 // Venice API endpoint
 const API_ENDPOINT = "https://api.venice.ai/api/v1/models?type=text";
 
-async function promptForApiKey(): Promise<string | null> {
-  const rl = readline.createInterface({
-    input: process.stdin,
-    output: process.stdout,
-  });
-
-  return new Promise((resolve) => {
-    rl.question(
-      "Enter Venice API key to include alpha models (or press Enter to skip): ",
-      (answer) => {
-        rl.close();
-        const trimmed = answer.trim();
-        resolve(trimmed.length > 0 ? trimmed : null);
-      },
-    );
-  });
-}
-
 // Zod schemas for API response validation
 const Capabilities = z
   .object({
@@ -69,6 +50,7 @@ const ModelSpec = z
   .object({
     pricing: Pricing.optional(),
     availableContextTokens: z.number(),
+    maxCompletionTokens: z.number().optional(),
     capabilities: Capabilities,
     constraints: z.any().optional(),
     name: z.string(),
@@ -257,11 +239,7 @@ function mergeModel(
   const caps = spec.capabilities;
 
   const contextTokens = spec.availableContextTokens;
-  const proposedOutputTokens = Math.floor(contextTokens / 4);
-  const outputTokens =
-    existing?.limit?.output !== undefined && existing.limit.output < proposedOutputTokens
-      ? existing.limit.output
-      : proposedOutputTokens
+  const outputTokens = spec.maxCompletionTokens ?? Math.floor(contextTokens / 4);
 
   const openWeights = spec.modelSource
     ? spec.modelSource.toLowerCase().includes("huggingface")
@@ -487,7 +465,7 @@ async function main() {
     "models",
   );
 
-  // Check for API key from CLI argument, environment, or prompt
+  // Check for API key from CLI argument or environment variable
   let apiKey: string | null = null;
 
   // Check CLI args for --api-key=xxx or --api-key xxx
@@ -506,11 +484,6 @@ async function main() {
     apiKey = process.env.VENICE_API_KEY ?? null;
   }
 
-  // Prompt if still no key
-  if (!apiKey) {
-    apiKey = await promptForApiKey();
-  }
-
   const includeAlpha = apiKey !== null;
 
   if (dryRun) {

packages/core/src/family.ts

@@ -128,9 +128,6 @@ export const ModelFamilyValues = [
   "solar-mini",
   "solar-pro",
 
-  // Exaone
-  "exaone",
-
   // Step (StepFun)
   "step",
 

providers/azure-cognitive-services/models/gpt-5.3-codex.toml

@@ -0,0 +1,23 @@
+name = "GPT-5.3 Codex"
+family = "gpt-codex"
+release_date = "2026-02-24"
+last_updated = "2026-02-24"
+attachment = false
+reasoning = true
+temperature = false
+knowledge = "2025-08-31"
+tool_call = true
+open_weights = false
+
+[cost]
+input = 1.75
+output = 14.00
+cache_read = 0.175
+
+[limit]
+context = 400_000
+output = 128_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/azure/models/gpt-5.3-codex.toml

@@ -0,0 +1,23 @@
+name = "GPT-5.3 Codex"
+family = "gpt-codex"
+release_date = "2026-02-24"
+last_updated = "2026-02-24"
+attachment = false
+reasoning = true
+temperature = false
+knowledge = "2025-08-31"
+tool_call = true
+open_weights = false
+
+[cost]
+input = 1.75
+output = 14.00
+cache_read = 0.175
+
+[limit]
+context = 400_000
+output = 128_000
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/friendli/models/LGAI-EXAONE/EXAONE-4.0.1-32B.toml

@@ -0,0 +1,25 @@
+name = "MiniMax M2.5"
+family = "minimax"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+temperature = true
+release_date = "2026-02-12"
+last_updated = "2026-02-12"
+open_weights = true
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.3
+output = 1.2
+
+[limit]
+context = 196_608
+output = 196_608
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/friendli/models/LGAI-EXAONE/K-EXAONE-236B-A23B.toml

@@ -15,7 +15,7 @@ field = "reasoning_content"
 [cost]
 input = 0.6
 output = 3.0
-cache_read = 0.08
+cache_read = 0.1
 
 [limit]
 context = 262_144

providers/friendli/models/MiniMaxAI/MiniMax-M2.5.toml

@@ -12,7 +12,7 @@ open_weights = true
 [cost]
 input = 0.3
 output = 1.2
-cache_read = 0.06
+cache_read = 0.03
 
 [limit]
 context = 204_800