@lobehub/chat
Version:
Lobe Chat - an open-source, high-performance chatbot framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.
72 lines (68 loc) • 1.92 kB
text/typescript
import { AIChatModelCard } from '@/types/aiModel';
const minimaxChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
search: true,
vision: true,
},
contextWindowTokens: 1_000_192,
description:
'在 MiniMax-01系列模型中,我们做了大胆创新:首次大规模实现线性注意力机制,传统 Transformer架构不再是唯一的选择。这个模型的参数量高达4560亿,其中单次激活459亿。模型综合性能比肩海外顶尖模型,同时能够高效处理全球最长400万token的上下文,是GPT-4o的32倍,Claude-3.5-Sonnet的20倍。',
displayName: 'MiniMax-Text-01',
enabled: true,
id: 'MiniMax-Text-01',
maxOutput: 1_000_192,
pricing: {
currency: 'CNY',
input: 1,
output: 8,
},
releasedAt: '2025-01-15',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
vision: true,
},
contextWindowTokens: 245_760,
description: '适用于广泛的自然语言处理任务,包括文本生成、对话系统等。',
displayName: 'abab6.5s',
enabled: true,
id: 'abab6.5s-chat',
maxOutput: 245_760,
pricing: {
currency: 'CNY',
input: 1,
output: 1,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 64_000,
description:
'DeepSeek 推出的推理模型。在输出最终回答之前,模型会先输出一段思维链内容,以提升最终答案的准确性。',
displayName: 'DeepSeek R1',
id: 'DeepSeek-R1',
maxOutput: 64_000,
pricing: {
currency: 'CNY',
input: 4,
output: 16,
},
type: 'chat',
},
];
export const allModels = [...minimaxChatModels];
export default allModels;