Abliterated Model Large V2: AI 大语言模型

通过 API 调用Abliterated Model Large V2。1M-token context, streaming and pay-per-token pricing. 可在 Muapi playground 免费试用。 Abliterated Model Large V2 is the second revision of the large abliterated reasoning model, with a 1M-token context window and tool-calling support. Muapi exposes it through a single REST endpoint with async polling and Server-Sent Events streaming, so you can drop it into chat apps, agents and workflows without hosting GPUs. It is best suited to large-context reasoning and tool-using agents. 按 token 计费：$6/M 输入 token，$10/M output tokens, ，无需订阅。 See also Abliterated Model, Abliterated Model Large. Browse the full lineup on the Abliterated Model API page or the abliterated LLM API hub; related families: GLM Abliterated API and Gemma Abliterated API. 'Abliterated' and 'derestricted' describe modifications that reduce the model's refusal behavior. They do not guarantee unrestricted output, and you remain responsible for complying with applicable laws and the Muapi terms of use.

交互式模型控制

通过 API 调用Abliterated Model Large V2。1M-token context, streaming and pay-per-token pricing. 可在 Muapi playground 免费试用。

📝

概览

关于此模型

Abliterated Model Large V2 is the second revision of the large abliterated reasoning model, with a 1M-token context window and tool-calling support. Muapi exposes it through a single REST endpoint with async polling and Server-Sent Events streaming, so you can drop it into chat apps, agents and workflows without hosting GPUs. It is best suited to large-context reasoning and tool-using agents. 按 token 计费：$6/M 输入 token，$10/M output tokens, ，无需订阅。 See also Abliterated Model, Abliterated Model Large. Browse the full lineup on the Abliterated Model API page or the abliterated LLM API hub; related families: GLM Abliterated API and Gemma Abliterated API. 'Abliterated' and 'derestricted' describe modifications that reduce the model's refusal behavior. They do not guarantee unrestricted output, and you remain responsible for complying with applicable laws and the Muapi terms of use.

1Chat and assistants: Build conversational apps on Abliterated Model Large V2 for large-context reasoning and tool-using agents.

2Low-refusal experimentation: Compare refusal behavior against the base model in your own evaluations, within Muapi's usage policy.

3Agents and workflows: 通过 API 调用the model from Muapi workflow nodes and the [abliterated LLM API](/abliterated-llm-api) hub examples.

💰

价格与价值

成本分析

提供商	费用	备注
muapiapp	$6/M input tokens and $10/M output tokens	按 token 计费，按实际用量结算，请求开始时预留少量最低额度。
Fal.ai	暂不可用	Abliterated Model Large V2 is not offered as a hosted endpoint.
Replicate	暂不可用	Abliterated Model Large V2 is not offered as a hosted endpoint.

muapiapp$6/M input tokens and $10/M output tokens

按 token 计费，按实际用量结算，请求开始时预留少量最低额度。

Fal.ai暂不可用

Abliterated Model Large V2 is not offered as a hosted endpoint.

Replicate暂不可用

Abliterated Model Large V2 is not offered as a hosted endpoint.

** 竞品价格根据相似模型架构和使用层级估算。

⚙️

技术详情

配置参数

参数	类型	描述	默认值
提示词	string	提供给模型的用户消息或指令。	`Summarize the trade-offs between a monolith and microservices in five bullet points.`
系统提示词	string	可选的系统级指令，用于引导模型行为。	`You are a concise, direct assistant.`
最大 Token 数	integer	可选的 maximum number of tokens to generate.	`8192`

提示词string

提供给模型的用户消息或指令。

默认值Summarize the trade-offs between a monolith and microservices in five bullet points.

系统提示词string

可选的系统级指令，用于引导模型行为。

默认值You are a concise, direct assistant.

最大 Token 数integer

可选的 maximum number of tokens to generate.

默认值8192

📖

实施指南

开发者文档

使用方法 Abliterated Model Large V2 API

获取 API 密钥: 在 Muapi 控制台创建密钥。
发送提示词: POST prompt (plus optional system_prompt and max_tokens, up to 32768) to /api/v1/abliterated-model-large-v2.

curl -X POST https://api.muapi.ai/api/v1/abliterated-model-large-v2 \
  -H "x-api-key: $MUAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Explain how attention works in transformers."}'

轮询或流式接收: Poll GET /api/v1/predictions/{request_id}/result until status is completed, or stream tokens live from /api/v1/abliterated-model-large-v2/stream.

curl -X POST https://api.muapi.ai/api/v1/abliterated-model-large-v2/stream \
  -H "x-api-key: $MUAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Explain how attention works in transformers."}'

查看结果: The generated text is returned in outputs. Usage is billed on actual token counts.

❓

常见问题

常见问答

什么是Abliterated Model Large V2?

Abliterated Model Large V2 is the second revision of the large abliterated reasoning model, with a 1M-token context window and tool-calling support.

abliterated（消减拒答）是什么意思？

Abliterated 或 derestricted 模型经过修改以减少拒答行为。这会降低拒答概率，但无法完全消除拒答，也不保证输出不受限制。

如何计费：Abliterated Model Large V2

Billing is per token: $6/M 输入 token，$10/M 输出 token。 A small minimum is reserved when the request starts and reconciled to actual usage.

是否Abliterated Model Large V2支持流式输出？

支持。 Stream tokens live from `/api/v1/abliterated-model-large-v2/stream` using Server-Sent Events, or poll the async endpoint.

上下文窗口有多大？

The context window is about 1M tokens. Prompt and output together must fit inside it; `max_tokens` is capped at 32768. The endpoint accepts text input only.

Is output guaranteed to be unrestricted?

不支持。 Abliterated models are less likely to refuse, but behavior varies by prompt and quality can differ from the base model. Evaluate it for your use case and follow applicable laws.