Merge branch 'enricoros:v2-dev' into v2-dev

2026-05-10 21:50:14 -07:00 · 2025-01-07 10:40:09 +08:00
parent 9411cd4f9b 271e5f3fc9
commit e971db1c60
5 changed files with 382 additions and 96 deletions
@@ -356,7 +356,13 @@ export function prettyShortChatModelName(model: string | undefined): string {
  if (model.includes(':'))
    return model.replace(':latest', '').replaceAll(':', ' ');
  // [xAI]
-  if (model.includes('grok-beta')) return 'Grok Beta';
+  if (model.includes('grok-')) {
+    if (model.includes('grok-3')) return 'Grok 3';
+    if (model.includes('grok-2-vision')) return 'Grok 2 Vision';
+    if (model.includes('grok-2')) return 'Grok 2';
+    if (model.includes('grok-beta')) return 'Grok Beta';
+    if (model.includes('grok-vision-beta')) return 'Grok Vision Beta';
+  }
  return model;
 }

@@ -6,7 +6,6 @@ import { _knownOpenAIChatModels } from './openai.models';
 import { wireGroqModelsListOutputSchema } from '../groq.wiretypes';
 import { wireOpenPipeModelOutputSchema } from '../openpipe.wiretypes';
 import { wireOpenrouterModelsListOutputSchema } from '../openrouter.wiretypes';
-import { wireTogetherAIListOutputSchema } from '../togetherai.wiretypes';


 export function azureModelToModelDescription(azureDeploymentRef: string, openAIModelIdBase: string, modelCreated: number, modelUpdated?: number): ModelDescriptionSchema {
@@ -338,78 +337,6 @@ export function openRouterModelToModelDescription(wireModel: object): ModelDescr
 }


-// [Together AI]
-
-const _knownTogetherAIChatModels: ManualMappings = [
-  {
-    idPrefix: 'NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO',
-    label: 'Nous Hermes 2 - Mixtral 8x7B-DPO',
-    description: 'Nous Hermes 2 Mixtral 7bx8 DPO is the new flagship Nous Research model trained over the Mixtral 7bx8 MoE LLM. The model was trained on over 1,000,000 entries of primarily GPT-4 generated data, as well as other high quality data from open datasets across the AI landscape, achieving state of the art performance on a variety of tasks.',
-    contextWindow: 32768,
-    interfaces: [LLM_IF_OAI_Chat],
-  },
-  {
-    idPrefix: 'NousResearch/Nous-Hermes-2-Mixtral-8x7B-SFT',
-    label: 'Nous Hermes 2 - Mixtral 8x7B-SFT',
-    description: 'Nous Hermes 2 Mixtral 7bx8 SFT is the new flagship Nous Research model trained over the Mixtral 7bx8 MoE LLM. The model was trained on over 1,000,000 entries of primarily GPT-4 generated data, as well as other high quality data from open datasets across the AI landscape, achieving state of the art performance on a variety of tasks.',
-    contextWindow: 32768,
-    interfaces: [LLM_IF_OAI_Chat],
-  },
-  {
-    idPrefix: 'mistralai/Mixtral-8x7B-Instruct-v0.1',
-    label: 'Mixtral-8x7B Instruct',
-    description: 'The Mixtral-8x7B Large Language Model (LLM) is a pretrained generative Sparse Mixture of Experts.',
-    contextWindow: 32768,
-    interfaces: [LLM_IF_OAI_Chat],
-  },
-  {
-    idPrefix: 'mistralai/Mistral-7B-Instruct-v0.2',
-    label: 'Mistral (7B) Instruct v0.2',
-    description: 'The Mistral-7B-Instruct-v0.2 Large Language Model (LLM) is an improved instruct fine-tuned version of Mistral-7B-Instruct-v0.1.',
-    contextWindow: 32768,
-    interfaces: [LLM_IF_OAI_Chat],
-  },
-  {
-    idPrefix: 'NousResearch/Nous-Hermes-2-Yi-34B',
-    label: 'Nous Hermes-2 Yi (34B)',
-    description: 'Nous Hermes 2 - Yi-34B is a state of the art Yi Fine-tune',
-    contextWindow: 4097,
-    interfaces: [LLM_IF_OAI_Chat],
-  },
-] as const;
-
-export function togetherAIModelsToModelDescriptions(wireModels: unknown): ModelDescriptionSchema[] {
-
-  function togetherAIModelToModelDescription(model: { id: string, created: number }) {
-    return fromManualMapping(_knownTogetherAIChatModels, model.id, model.created, undefined, {
-      idPrefix: model.id,
-      label: model.id.replaceAll('/', ' · ').replaceAll(/[_-]/g, ' '),
-      description: 'New Togehter AI Model',
-      contextWindow: null, // unknown
-      interfaces: [LLM_IF_OAI_Chat], // assume..
-      hidden: true,
-    });
-  }
-
-  function togetherAIModelsSort(a: ModelDescriptionSchema, b: ModelDescriptionSchema): number {
-    if (a.hidden && !b.hidden)
-      return 1;
-    if (!a.hidden && b.hidden)
-      return -1;
-    if (a.created !== b.created)
-      return (b.created || 0) - (a.created || 0);
-    return a.id.localeCompare(b.id);
-  }
-
-  return wireTogetherAIListOutputSchema.parse(wireModels)
-    .map(togetherAIModelToModelDescription)
-    .sort(togetherAIModelsSort);
-}
-
-
-// Perplexity
-
-
 // Groq - https://console.groq.com/docs/models

 const _knownGroqModels: ManualMappings = [
@@ -0,0 +1,283 @@
+import { LLM_IF_OAI_Chat } from '~/common/stores/llms/llms.types';
+
+import type { ModelDescriptionSchema } from '../../llm.server.types';
+import { fromManualMapping, ManualMappings } from './models.data';
+import { wireTogetherAIListOutputSchema } from '../togetherai.wiretypes';
+
+
+const _knownTogetherAIChatModels: ManualMappings = [
+  {
+    idPrefix: 'meta-llama/Llama-3.3-70B-Instruct-Turbo',
+    label: 'Llama 3.3 70B Instruct Turbo',
+    description: 'Llama 3.3 70B Instruct Turbo is an advanced model from Meta with a context length of 131072 tokens, using FP8 quantization.',
+    contextWindow: 131072,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo',
+    label: 'Llama 3.1 8B Instruct Turbo',
+    description: 'Llama 3.1 8B Instruct Turbo is an advanced model from Meta with a context length of 131072 tokens, using FP8 quantization.',
+    contextWindow: 131072,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo',
+    label: 'Llama 3.1 70B Instruct Turbo',
+    description: 'Llama 3.1 70B Instruct Turbo is an advanced model from Meta with a context length of 131072 tokens, using FP8 quantization.',
+    contextWindow: 131072,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo',
+    label: 'Llama 3.1 405B Instruct Turbo',
+    description: 'Llama 3.1 405B Instruct Turbo is a large-scale model from Meta with a context length of 130815 tokens, using FP8 quantization.',
+    contextWindow: 130815,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'meta-llama/Meta-Llama-3-8B-Instruct-Turbo',
+    label: 'Llama 3 8B Instruct Turbo',
+    description: 'Llama 3 8B Instruct Turbo is an advanced model from Meta with a context length of 8192 tokens, using FP8 quantization.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'meta-llama/Meta-Llama-3-70B-Instruct-Turbo',
+    label: 'Llama 3 70B Instruct Turbo',
+    description: 'Llama 3 70B Instruct Turbo is an advanced model from Meta with a context length of 8192 tokens, using FP8 quantization.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'meta-llama/Llama-3.2-3B-Instruct-Turbo',
+    label: 'Llama 3.2 3B Instruct Turbo',
+    description: 'Llama 3.2 3B Instruct Turbo is a model from Meta with a context length of 131072 tokens, using FP16 precision.',
+    contextWindow: 131072,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'meta-llama/Meta-Llama-3-8B-Instruct-Lite',
+    label: 'Llama 3 8B Instruct Lite',
+    description: 'Llama 3 8B Instruct Lite is an efficient model from Meta with a context length of 8192 tokens, using INT4 quantization.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'meta-llama/Meta-Llama-3-70B-Instruct-Lite',
+    label: 'Llama 3 70B Instruct Lite',
+    description: 'Llama 3 70B Instruct Lite is an efficient model from Meta with a context length of 8192 tokens, using INT4 quantization.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'meta-llama/Llama-3-8b-chat-hf',
+    label: 'Llama 3 8B Instruct Reference',
+    description: 'Llama 3 8B Instruct Reference is a base model from Meta with a context length of 8192 tokens.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'meta-llama/Llama-3-70b-chat-hf',
+    label: 'Llama 3 70B Instruct Reference',
+    description: 'Llama 3 70B Instruct Reference is a base model from Meta with a context length of 8192 tokens.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'nvidia/Llama-3.1-Nemotron-70B-Instruct-HF',
+    label: 'Llama 3.1 Nemotron 70B',
+    description: 'Llama 3.1 Nemotron 70B is a model from Nvidia with a context length of 32768 tokens, using FP16 precision.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'Qwen/Qwen2.5-Coder-32B-Instruct',
+    label: 'Qwen 2.5 Coder 32B Instruct',
+    description: 'Qwen 2.5 Coder 32B Instruct is a model from Qwen optimized for code tasks, with a context length of 32768 tokens.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'Qwen/QwQ-32B-Preview',
+    label: 'QwQ-32B-Preview',
+    description: 'QwQ-32B-Preview is a preview model from Qwen with a context length of 32768 tokens.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'microsoft/WizardLM-2-8x22B',
+    label: 'WizardLM-2 8x22B',
+    description: 'WizardLM-2 8x22B is a model from Microsoft with a context length of 65536 tokens.',
+    contextWindow: 65536,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'google/gemma-2-27b-it',
+    label: 'Gemma 2 27B',
+    description: 'Gemma 2 27B is a model from Google with a context length of 8192 tokens.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'google/gemma-2-9b-it',
+    label: 'Gemma 2 9B',
+    description: 'Gemma 2 9B is a model from Google with a context length of 8192 tokens.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'databricks/dbrx-instruct',
+    label: 'DBRX Instruct',
+    description: 'DBRX Instruct is a model from Databricks with a context length of 32768 tokens.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'deepseek-ai/deepseek-llm-67b-chat',
+    label: 'DeepSeek LLM Chat (67B)',
+    description: 'DeepSeek LLM Chat (67B) is a model from DeepSeek AI with a context length of 4096 tokens.',
+    contextWindow: 4096,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'deepseek-ai/DeepSeek-V3',
+    label: 'DeepSeek-V3',
+    description: 'DeepSeek-V3 is a model from DeepSeek AI with a context length of 131072 tokens, using FP8 quantization.',
+    contextWindow: 131072,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'google/gemma-2b-it',
+    label: 'Gemma Instruct (2B)',
+    description: 'Gemma Instruct (2B) is a model from Google with a context length of 8192 tokens.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'Gryphe/MythoMax-L2-13b',
+    label: 'MythoMax-L2 (13B)',
+    description: 'MythoMax-L2 (13B) is a model from Gryphe with a context length of 4096 tokens.',
+    contextWindow: 4096,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'meta-llama/Llama-2-13b-chat-hf',
+    label: 'LLaMA-2 Chat (13B)',
+    description: 'LLaMA-2 Chat (13B) is a model from Meta with a context length of 4096 tokens.',
+    contextWindow: 4096,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'mistralai/Mistral-7B-Instruct-v0.1',
+    label: 'Mistral (7B) Instruct',
+    description: 'Mistral (7B) Instruct v0.1 is the initial instruct fine-tuned version of the Mistral 7B model.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'mistralai/Mistral-7B-Instruct-v0.2',
+    label: 'Mistral (7B) Instruct v0.2',
+    description: 'Mistral (7B) Instruct v0.2 is an improved version with extended context length.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'mistralai/Mistral-7B-Instruct-v0.3',
+    label: 'Mistral (7B) Instruct v0.3',
+    description: 'Mistral (7B) Instruct v0.3 is the latest version of Mistral 7B Instruct.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'mistralai/Mixtral-8x7B-Instruct-v0.1',
+    label: 'Mixtral-8x7B Instruct (46.7B)',
+    description: 'Mixtral-8x7B Instruct is a model from Mistral AI with a context length of 32768 tokens.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'mistralai/Mixtral-8x22B-Instruct-v0.1',
+    label: 'Mixtral-8x22B Instruct (141B)',
+    description: 'Mixtral-8x22B Instruct is a larger model from Mistral AI with a context length of 65536 tokens.',
+    contextWindow: 65536,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO',
+    label: 'Nous Hermes 2 - Mixtral 8x7B-DPO (46.7B)',
+    description: 'Nous Hermes 2 - Mixtral 8x7B-DPO is a model from NousResearch with a context length of 32768 tokens.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'NousResearch/Nous-Hermes-2-Mixtral-8x7B-SFT',
+    label: 'Nous Hermes 2 - Mixtral 8x7B-SFT',
+    description: 'Nous Hermes 2 Mixtral 8x7B-SFT is a model from Nous Research trained over the Mixtral 8x7B MoE LLM.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'NousResearch/Nous-Hermes-2-Yi-34B',
+    label: 'Nous Hermes-2 Yi (34B)',
+    description: 'Nous Hermes 2 - Yi-34B is a state-of-the-art model from Nous Research.',
+    contextWindow: 4097,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'Qwen/Qwen2.5-7B-Instruct-Turbo',
+    label: 'Qwen 2.5 7B Instruct Turbo',
+    description: 'Qwen 2.5 7B Instruct Turbo is a model from Qwen with a context length of 32768 tokens, using FP8 quantization.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'Qwen/Qwen2.5-72B-Instruct-Turbo',
+    label: 'Qwen 2.5 72B Instruct Turbo',
+    description: 'Qwen 2.5 72B Instruct Turbo is a model from Qwen with a context length of 32768 tokens, using FP8 quantization.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'Qwen/Qwen2-72B-Instruct',
+    label: 'Qwen 2 Instruct (72B)',
+    description: 'Qwen 2 Instruct (72B) is a model from Qwen with a context length of 32768 tokens.',
+    contextWindow: 32768,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+  {
+    idPrefix: 'upstage/SOLAR-10.7B-Instruct-v1.0',
+    label: 'Upstage SOLAR Instruct v1 (11B)',
+    description: 'Upstage SOLAR Instruct v1 is a model from Upstage with a context length of 4096 tokens.',
+    contextWindow: 4096,
+    interfaces: [LLM_IF_OAI_Chat],
+  },
+] as const;
+
+
+export function togetherAIModelsToModelDescriptions(wireModels: unknown): ModelDescriptionSchema[] {
+
+  function togetherAIModelToModelDescription(model: { id: string, created: number }) {
+    return fromManualMapping(_knownTogetherAIChatModels, model.id, model.created, undefined, {
+      idPrefix: model.id,
+      label: model.id.replaceAll('/', ' · ').replaceAll(/[_-]/g, ' '),
+      description: 'New Together AI Model',
+      contextWindow: null, // unknown
+      interfaces: [LLM_IF_OAI_Chat], // assume
+      hidden: true,
+    });
+  }
+
+  function togetherAIModelsSort(a: ModelDescriptionSchema, b: ModelDescriptionSchema): number {
+    if (a.hidden && !b.hidden)
+      return 1;
+    if (!a.hidden && b.hidden)
+      return -1;
+    if (a.created !== b.created)
+      return (b.created || 0) - (a.created || 0);
+    return a.id.localeCompare(b.id);
+  }
+
+  return wireTogetherAIListOutputSchema.parse(wireModels)
+    .map(togetherAIModelToModelDescription)
+    .sort(togetherAIModelsSort);
+}
@@ -5,18 +5,47 @@ import { fetchJsonOrTRPCThrow } from '~/server/trpc/trpc.router.fetchers';
 import { LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, LLM_IF_OAI_Vision } from '~/common/stores/llms/llms.types';

 import type { ModelDescriptionSchema } from '../../llm.server.types';
-import { fromManualMapping, ManualMappings } from './models.data';
+import { fromManualMapping, ManualMapping, ManualMappings } from './models.data';
 import { openAIAccess, OpenAIAccessSchema } from '../openai.router';


 // Known xAI Models - Manual Mappings
 // List on: https://console.x.ai/team/_TEAM_ID_/models
 const _knownXAIChatModels: ManualMappings = [
+
+  {
+    idPrefix: 'grok-2-vision-1212',
+    label: `Grok 2 Vision (1212)`,
+    description: 'xAI model grok-2-vision-1212 with image and text input capabilities. Supports text generation with an 8,192 token context window.',
+    contextWindow: 8192,
+    maxCompletionTokens: undefined,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, LLM_IF_OAI_Vision],
+    chatPrice: { input: 2, output: 10 },
+  },
+  {
+    idPrefix: 'grok-2-1212',
+    label: `Grok 2 (1212)`,
+    description: 'xAI model grok-2-1212 with text input capabilities. Supports text generation with a 131,072 token context window.',
+    contextWindow: 131072,
+    maxCompletionTokens: undefined,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
+    chatPrice: { input: 2, output: 10 },
+  },
+
+  {
+    idPrefix: 'grok-vision-beta',
+    label: `Grok Vision Beta`,
+    description: 'xAI model grok-vision-beta with image and text input capabilities. Supports text generation with an 8,192 token context window.',
+    contextWindow: 8192,
+    maxCompletionTokens: undefined,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, LLM_IF_OAI_Vision],
+    chatPrice: { input: 5, output: 15 },
+  },
  {
    idPrefix: 'grok-beta',
    label: `Grok Beta`,
-    description: 'xAI\'s flagship model with real-time knowledge from the X platform. Supports text generation with a 131K token context window.',
-    contextWindow: 131072,  // 131,072 tokens as shown in the Context column
+    description: 'xAI\'s flagship model with real-time knowledge from the X platform. Supports text generation with a 131,072 token context window.',
+    contextWindow: 131072,
    maxCompletionTokens: 16384,
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
    chatPrice: { input: 5, output: 15 },
@@ -33,23 +62,62 @@ export async function xaiModelDescriptions(access: OpenAIAccessSchema): Promise<

  const xaiModels = wireXAIModelsListSchema.parse(modelsResponse);

-  return xaiModels.models.map(model => fromManualMapping(_knownXAIChatModels, model.id, model.created, undefined, {
-    idPrefix: model.id,
-    label: `${model.id} ${model.version || ''}`, // {{Created}}`,
-    description: `xAI model ${model.id}`,
-    contextWindow: 16384,
-    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, ...(model.input_modalities?.includes('image') ? [LLM_IF_OAI_Vision] : [])],
-    ...(model.prompt_text_token_price && model.completion_text_token_price && {
-      chatPrice: {
-        input: model.prompt_text_token_price / 10000, // FIXME: SCALE UNKNOWN for now
-        output: model.completion_text_token_price / 10000,
-      },
-    }),
-  }));
+  return xaiModels.models.reduce((acc, xm) => {
+
+    // Fallback for unknown models
+    const unknownModelFallback: ManualMapping = {
+      idPrefix: xm.id,
+      label: `${xm.id}${xm.version ? ' ' + xm.version : ''}`,
+      description: `xAI model ${xm.id}`,
+      contextWindow: 16384,
+      interfaces: [
+        LLM_IF_OAI_Chat,
+        LLM_IF_OAI_Fn,
+        ...(xm.input_modalities?.includes('image') ? [LLM_IF_OAI_Vision] : []),
+      ],
+      ...(xm.prompt_text_token_price != null && xm.completion_text_token_price != null && {
+        chatPrice: {
+          input: xm.prompt_text_token_price / 10000, // Scaling factor applied as per API data
+          output: xm.completion_text_token_price / 10000,
+        },
+      }),
+    };
+
+    // xAI model description
+    const modelDescription = fromManualMapping(_knownXAIChatModels, xm.id, xm.created, undefined, unknownModelFallback);
+    acc.push(modelDescription);
+
+    // NOTE: disabled, as this is not useful
+    // if there are aliases, add them as 'symlinked' models
+    // if (xm.aliases?.length) {
+    //   xm.aliases.forEach((alias) => {
+    //     const aliasedModel = fromManualMapping([{
+    //       idPrefix: alias,
+    //       label: alias,
+    //       symLink: xm.id,
+    //       description: `xAI model ${alias}`,
+    //       contextWindow: 16384,
+    //       interfaces: unknownModelFallback.interfaces,
+    //     }], alias, xm.created, xm.updated, unknownModelFallback);
+    //     acc.push(aliasedModel);
+    //   });
+    // }
+
+    return acc;
+  }, [] as ModelDescriptionSchema[]);
 }

+// manual sort order
+const _xaiLabelStartsWithOrder = ['Grok 3', 'Grok 2', 'Grok'];
+
 export function xaiModelSort(a: ModelDescriptionSchema, b: ModelDescriptionSchema): number {
-  return b.label.localeCompare(a.label);
+  const aStartsWith = _xaiLabelStartsWithOrder.findIndex((prefix) => a.label.startsWith(prefix));
+  const bStartsWith = _xaiLabelStartsWithOrder.findIndex((prefix) => b.label.startsWith(prefix));
+
+  if (aStartsWith !== bStartsWith)
+    return aStartsWith - bStartsWith;
+
+  return a.label.localeCompare(b.label);
 }


@@ -71,15 +139,16 @@ export const wireXAIModelSchema = z.object({
  version: z.string().optional(),

  // modalities
-  input_modalities: z.array(z.string()),    // relaxing it
-  output_modalities: z.array(z.string()),   // relaxing it
-  // input_modalities: z.array(z.enum(['text'])),
-  // output_modalities: z.array(z.enum(['text'])),
+  input_modalities: z.array(z.string()),    // 'text', 'image', etc.
+  output_modalities: z.array(z.string()),   // 'text', 'image', etc.

  // pricing - FIXME: SCALE UNKNOWN for now
  prompt_text_token_price: z.number().optional(),
  prompt_image_token_price: z.number().optional(),
  completion_text_token_price: z.number().optional(),
+
+  // Aliases for models
+  aliases: z.array(z.string()).optional(),
 });

 export const wireXAIModelsListSchema = z.object({
@@ -13,11 +13,12 @@ import { fixupHost } from '~/common/util/urlUtils';
 import { OpenAIWire_API_Images_Generations, OpenAIWire_API_Models_List, OpenAIWire_API_Moderations_Create } from '~/modules/aix/server/dispatch/wiretypes/openai.wiretypes';

 import { ListModelsResponse_schema, ModelDescriptionSchema } from '../llm.server.types';
-import { azureModelToModelDescription, groqModelSortFn, groqModelToModelDescription, lmStudioModelToModelDescription, localAIModelToModelDescription, openPipeModelDescriptions, openPipeModelSort, openPipeModelToModelDescriptions, openRouterModelFamilySortFn, openRouterModelToModelDescription, togetherAIModelsToModelDescriptions } from './models/models.data';
+import { azureModelToModelDescription, groqModelSortFn, groqModelToModelDescription, lmStudioModelToModelDescription, localAIModelToModelDescription, openPipeModelDescriptions, openPipeModelSort, openPipeModelToModelDescriptions, openRouterModelFamilySortFn, openRouterModelToModelDescription } from './models/models.data';
 import { deepseekModelFilter, deepseekModelToModelDescription } from './models/deepseek.models';
 import { mistralModelsSort, mistralModelToModelDescription } from './models/mistral.models';
 import { openAIModelFilter, openAIModelToModelDescription, openAISortModels } from './models/openai.models';
 import { perplexityAIModelDescriptions, perplexityAIModelSort } from './models/perplexity.models';
+import { togetherAIModelsToModelDescriptions } from './models/together.models';
 import { wilreLocalAIModelsApplyOutputSchema, wireLocalAIModelsAvailableOutputSchema, wireLocalAIModelsListOutputSchema } from './localai.wiretypes';
 import { xaiModelDescriptions, xaiModelSort } from './models/xai.models';