> ## Documentation Index
> Fetch the complete documentation index at: https://apidocs.mor.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Available Models

> Explore all models available through the Morpheus Inference API

# Available Models

The Morpheus Inference Marketplace provides access to a variety of open-source AI models. Models are hosted by providers in the decentralized marketplace, and availability may vary based on provider activity.

## Large Language Models (LLMs)

### Flagship Models

These are the most capable models available for complex tasks.

| Model                            | Context Window | Capabilities                      | Best For                                                        |
| -------------------------------- | -------------- | --------------------------------- | --------------------------------------------------------------- |
| `deepseek-v4-pro`                | 1M             | Code, Function Calling, Reasoning | Frontier reasoning, complex coding, ultra-long context analysis |
| `glm-5.1-non-thinking`           | 200K           | Code, Function Calling            | High-throughput inference, tool-use without extended thinking   |
| `minimax-m2.5`                   | 198K           | Code, Function Calling, Reasoning | AI agents, autonomous workflows, multi-step tool orchestration  |
| `qwen3-coder-480b-a35b-instruct` | 256K           | Code, Function Calling            | Code generation, programming                                    |
| `hermes-3-llama-3.1-405b`        | 128K           | —                                 | General purpose, instruction following                          |
| `gpt-oss-120b`                   | 128K           | Function Calling                  | GPT-style responses                                             |

### Reasoning Models

Models optimized for step-by-step thinking and complex problem solving.

| Model                          | Context Window | Capabilities                              | Best For                                                          |
| ------------------------------ | -------------- | ----------------------------------------- | ----------------------------------------------------------------- |
| `glm-5.2`                      | 200K           | Code, Function Calling, Reasoning         | Next-gen agentic engineering, advanced reasoning, complex systems |
| `glm-5.1`                      | 200K           | Code, Function Calling, Reasoning         | Agentic engineering, complex systems, next-gen reasoning          |
| `kimi-k2.7-code`               | 256K           | Code, Function Calling, Reasoning         | Code generation, agentic coding, parallel agent workflows         |
| `kimi-k2.6`                    | 256K           | Code, Function Calling, Reasoning, Vision | Math, visual reasoning, parallel agent workflows                  |
| `MiniMax-M2.7`                 | 198K           | Code, Function Calling, Reasoning         | AI agents, autonomous workflows, cost-efficient orchestration     |
| `glm-5`                        | 200K           | Code, Function Calling, Reasoning         | Agentic engineering, complex systems, long-horizon tasks          |
| `kimi-k2.5`                    | 256K           | Code, Function Calling, Reasoning, Vision | Math, visual reasoning, parallel agent workflows                  |
| `glm-4.7-thinking`             | 198K           | Function Calling, Reasoning               | Extended thinking, analysis                                       |
| `qwen3-235b`                   | 128K           | Function Calling                          | Complex reasoning, long documents                                 |
| `arcee-trinity-large-thinking` | 256K           | Code, Function Calling, Reasoning         | Agentic workflows, multi-step planning, tool orchestration        |
| `gemma-4-26b-a4b`              | 256K           | Reasoning, Vision                         | Efficient reasoning, low-latency inference, image analysis        |
| `gemma-4-31b`                  | 256K           | Reasoning, Vision                         | Math, science, coding, document parsing, visual reasoning         |

### Mid-Size Models

Balanced performance and speed for most use cases.

| Model               | Context Window | Capabilities             | Best For                                      |
| ------------------- | -------------- | ------------------------ | --------------------------------------------- |
| `llama-3.3-70b`     | 128K           | Function Calling         | General purpose, reliable                     |
| `qwen3-next-80b`    | 256K           | Function Calling         | Next-gen reasoning, long context              |
| `qwen35-35b-a3b`    | 256K           | Function Calling, Vision | General purpose, long context, image analysis |
| `mistral-31-24b`    | 128K           | Function Calling, Vision | Fast, efficient, image analysis               |
| `venice-uncensored` | 32K            | —                        | Uncensored, creative, roleplay                |

### Fast Models

Optimized for speed and low latency.

| Model               | Context Window | Capabilities                      | Best For                                                  |
| ------------------- | -------------- | --------------------------------- | --------------------------------------------------------- |
| `deepseek-v4-flash` | 1M             | Code, Function Calling, Reasoning | Fast frontier-tier reasoning, long-context tasks at speed |
| `glm-4.7-flash`     | 200K           | Function Calling, Reasoning       | Agentic coding, tool-use workflows, local deployment      |
| `qwen35-9b`         | 256K           | Function Calling, Vision          | Fast responses, image analysis, simple tasks              |
| `llama-3.2-3b`      | 128K           | Function Calling                  | Fastest responses, simple tasks                           |

## Proprietary, Anonymized Models

Frontier proprietary models served through anonymized providers on the Morpheus marketplace. These models are routed to closed-source frontier providers without exposing your identity to the upstream provider. Like all other models, each one supports web search by appending `:web` to the model name (e.g., `claude-fable-5:web`).

| Model                    | Best For                                   |
| ------------------------ | ------------------------------------------ |
| `claude-fable-5`         | Frontier agentic coding and reasoning      |
| `claude-opus-4.8`        | Advanced reasoning and code development    |
| `gemini-3.1-pro-preview` | Multimodal frontier reasoning              |
| `GPT-5.5-Pro`            | Maximum-capability frontier reasoning      |
| `GPT-5.5`                | Frontier chat, reasoning, and coding       |
| `grok-4.20`              | Fast, cost-efficient frontier intelligence |

See [Model Pricing](/documentation/models/pricing) for per-token rates for these models.

## Embeddings Models

For vector embeddings and semantic search.

| Model                   | Best For                              |
| ----------------------- | ------------------------------------- |
| `text-embedding-bge-m3` | Text embeddings, RAG, semantic search |

## Audio Models

### Text-to-Speech

| Model        | Best For                         |
| ------------ | -------------------------------- |
| `tts-kokoro` | Natural-sounding voice synthesis |

## Model Capabilities

<AccordionGroup>
  <Accordion title="Function Calling">
    Models with function calling can invoke tools and APIs. Use the `tools` parameter in your chat completion request to define available functions.

    **Supported models:** Most models except `venice-uncensored` and `hermes-3-llama-3.1-405b`. Includes `glm-5.2`, `glm-5.1`, `glm-5.1-non-thinking`, `kimi-k2.7-code`, `kimi-k2.6`, `MiniMax-M2.7`, and more.
  </Accordion>

  <Accordion title="Reasoning">
    Reasoning models support extended thinking and step-by-step problem solving. They're optimized for complex math, logic, and analytical tasks.

    **Supported models:** `glm-5.2`, `glm-5.1`, `glm-5`, `kimi-k2.7-code`, `kimi-k2.6`, `kimi-k2.5`, `glm-4.7-thinking`, `glm-4.7-flash`, `glm-4.7`, `minimax-m2.5`, `MiniMax-M2.7`, `arcee-trinity-large-thinking`, `gemma-4-26b-a4b`, `gemma-4-31b`, `deepseek-v4-pro`, `deepseek-v4-flash`
  </Accordion>

  <Accordion title="Vision">
    Vision-capable models can analyze images passed in the messages array.

    **Supported models:** `mistral-31-24b`, `kimi-k2.6`, `kimi-k2.5`, `gemma-4-26b-a4b`, `gemma-4-31b`, `qwen35-35b-a3b`, `qwen35-9b`
  </Accordion>

  <Accordion title="Web Search">
    Any model can be upgraded with real-time web search by appending `:web` to the model name. See the [Web Search](#web-search-with-web) section below for details.

    **Available for:** All models
  </Accordion>

  <Accordion title="Code Optimization">
    Models specifically optimized for code generation and programming tasks.

    **Supported models:** `qwen3-coder-480b-a35b-instruct`, `glm-5.2`, `glm-5.1`, `glm-5`, `kimi-k2.7-code`, `kimi-k2.6`, `MiniMax-M2.7`, `minimax-m2.5`, `arcee-trinity-large-thinking`, `deepseek-v4-pro`, `deepseek-v4-flash`
  </Accordion>
</AccordionGroup>

## Web Search with `:web`

Every model listed above can be enhanced with real-time web search capabilities — you don't need a separate model for it. Simply append `:web` to any model name, and the model will search the internet for current information before generating its response.

<Note>
  The model tables above only list base model names. To use any model **with web search**, just add `:web` to the end. For example, `llama-3.3-70b` becomes `llama-3.3-70b:web`. This works universally across every model in the Morpheus marketplace.
</Note>

| Base Model       | With Web Search      | What Changes                                   |
| ---------------- | -------------------- | ---------------------------------------------- |
| `glm-5.2`        | `glm-5.2:web`        | Adds real-time internet search to responses    |
| `glm-5.1`        | `glm-5.1:web`        | Adds real-time internet search to responses    |
| `kimi-k2.7-code` | `kimi-k2.7-code:web` | Combines advanced coding with current web data |
| `kimi-k2.6`      | `kimi-k2.6:web`      | Combines deep reasoning with current web data  |
| `MiniMax-M2.7`   | `MiniMax-M2.7:web`   | Adds web search to agent workflows             |
| `glm-5`          | `glm-5:web`          | Adds real-time internet search to responses    |
| `kimi-k2.5`      | `kimi-k2.5:web`      | Combines deep reasoning with current web data  |
| `glm-4.7-flash`  | `glm-4.7-flash:web`  | Adds web search to fast, efficient responses   |
| *any model*      | *model-name*`:web`   | Same pattern — works for all models            |

<Tip>
  **Verify exact model names** — including `:web` variants — by querying the [`/models` endpoint](/api-reference/models/all-models). The `:web` suffix is universal, but the base model name must match exactly what the API returns.
</Tip>

## Using Models

Specify the model ID in your API requests:

<Tabs>
  <Tab title="curl">
    ```bash theme={null}
    curl https://api.mor.org/api/v1/chat/completions \
      -H "Authorization: Bearer YOUR_API_KEY" \
      -H "Content-Type: application/json" \
      -d '{
        "model": "llama-3.3-70b",
        "messages": [
          {"role": "user", "content": "Hello!"}
        ]
      }'
    ```
  </Tab>

  <Tab title="Python">
    ```python theme={null}
    from openai import OpenAI

    client = OpenAI(
        api_key="YOUR_API_KEY",
        base_url="https://api.mor.org/api/v1"
    )

    response = client.chat.completions.create(
        model="llama-3.3-70b",
        messages=[
            {"role": "user", "content": "Hello!"}
        ]
    )
    ```
  </Tab>

  <Tab title="JavaScript">
    ```javascript theme={null}
    import OpenAI from "openai";

    const client = new OpenAI({
      apiKey: "YOUR_API_KEY",
      baseURL: "https://api.mor.org/api/v1",
    });

    const response = await client.chat.completions.create({
      model: "llama-3.3-70b",
      messages: [
        { role: "user", content: "Hello!" }
      ],
    });
    ```
  </Tab>
</Tabs>

## List Active Models

Query the API to see currently available models:

```bash theme={null}
curl https://api.mor.org/api/v1/models \
  -H "Authorization: Bearer YOUR_API_KEY"
```

<Tip>
  Model availability depends on active providers in the Morpheus Inference Marketplace. The API automatically routes your request to the highest-rated provider for your selected model.
</Tip>

## Model Selection Guide

<AccordionGroup>
  <Accordion title="Best for coding">
    * **`deepseek-v4-pro`** - Frontier reasoning + coding, 1M context for whole-repo analysis
    * **`qwen3-coder-480b-a35b-instruct`** - Top choice for code generation (256K context)
    * **`kimi-k2.7-code`** - Next-gen agentic coding, parallel agent workflows (256K context)
    * **`minimax-m2.5`** - SOTA agentic coding, full-stack development (198K context)
    * **`glm-5.2`** - Next-gen agentic engineering, multi-file systems design (200K context)
    * **`glm-5.1`** - Next-gen agentic engineering, multi-file systems design (200K context)
    * **`MiniMax-M2.7`** - Cost-efficient agentic coding, tool orchestration (198K context)
    * **`glm-5`** - Agentic engineering, multi-file systems design (200K context)
    * **`arcee-trinity-large-thinking`** - Agentic coding, multi-turn tool calling (256K context)
    * **`deepseek-v4-flash`** - Fast frontier-tier coding, 1M context at lower latency
    * **`llama-3.3-70b`** - Good balance of speed and quality
  </Accordion>

  <Accordion title="Best for long documents">
    * **`deepseek-v4-pro`** - 1M context, frontier reasoning over long documents
    * **`deepseek-v4-flash`** - 1M context at lower latency
    * **`qwen3-next-80b`** - 256K context window
    * **`kimi-k2.7-code`** - 256K context with agentic coding
    * **`kimi-k2.6`** - 256K context with multimodal reasoning
    * **`kimi-k2.5`** - 256K context with multimodal reasoning
    * **`qwen3-coder-480b-a35b-instruct`** - 256K context window
    * **`glm-5.2`** - 200K context, excellent at document analysis
    * **`glm-5.1`** - 200K context, excellent at document analysis
    * **`glm-5`** - 200K context, excellent at document analysis
    * **`minimax-m2.5`** - 198K context window
    * **`arcee-trinity-large-thinking`** - 256K context, agentic reasoning
    * **`gemma-4-31b`** - 256K context, multimodal reasoning
  </Accordion>

  <Accordion title="Best for speed">
    * **`deepseek-v4-flash`** - Fast frontier-tier with 1M context
    * **`glm-4.7-flash`** - 30B MoE (3B active), 200K context, runs on consumer GPUs
    * **`llama-3.2-3b`** - Very fast, 128K context
    * **`gemma-4-26b-a4b`** - MoE reasoning, near-4B inference speed (256K context)
    * **`qwen35-9b`** - Fast with vision support (256K context)
    * **`mistral-31-24b`** - Good speed with vision support
  </Accordion>

  <Accordion title="Best for reasoning">
    * **`deepseek-v4-pro`** - Frontier reasoning, 1M context for ultra-long chains
    * **`kimi-k2.7-code`** - Next-gen reasoning for code and agentic tasks, 256K context
    * **`kimi-k2.6`** - Next-gen math/logic benchmarks, multimodal, 256K context
    * **`kimi-k2.5`** - Top math/logic benchmarks (AIME 96%), multimodal, 256K context
    * **`glm-5.2`** - Next-gen agentic engineering, systems reasoning, 200K context
    * **`glm-5.1`** - Next-gen agentic engineering, systems reasoning, 200K context
    * **`glm-5`** - Agentic engineering, systems reasoning, 200K context
    * **`MiniMax-M2.7`** - Cost-efficient reasoning, agent workflows, 198K context
    * **`deepseek-v4-flash`** - Fast frontier reasoning, 1M context at lower latency
    * **`glm-4.7-thinking`** - Extended thinking mode, 198K context
    * **`qwen3-235b`** - Complex analysis, 128K context
    * **`arcee-trinity-large-thinking`** - Agentic reasoning, multi-step planning, 256K context
    * **`gemma-4-31b`** - Math, science, visual reasoning, 256K context
  </Accordion>

  <Accordion title="Best for AI agents">
    * **`minimax-m2.5`** - Purpose-built for agents, 80.2% SWE-Bench, multi-step tool orchestration
    * **`MiniMax-M2.7`** - Next-gen agents, cost-efficient orchestration, 198K context
    * **`glm-5.2`** - Next-gen long-horizon agentic tasks, advanced reasoning, 200K context
    * **`glm-5.1`** - Next-gen long-horizon agentic tasks, advanced reasoning, 200K context
    * **`glm-5`** - Long-horizon agentic tasks, #1 open-source on Vending Bench 2
    * **`kimi-k2.7-code`** - Agent Swarm coding, parallel sub-agent workflows, 256K context
    * **`kimi-k2.6`** - Agent Swarm with up to 100 parallel sub-agents
    * **`kimi-k2.5`** - Agent Swarm with up to 100 parallel sub-agents
    * **`arcee-trinity-large-thinking`** - Purpose-built for long-horizon agents, multi-turn tool calling
    * **`glm-4.7-flash`** - Lightweight agentic coding, efficient tool-use workflows
  </Accordion>

  <Accordion title="Best for uncensored/creative">
    * **`venice-uncensored`** - Minimal content restrictions, roleplay
  </Accordion>
</AccordionGroup>

## Next Steps

<CardGroup cols={2}>
  <Card title="Quickstart" icon="rocket" href="/quickstart">
    Get started making your first API call.
  </Card>

  <Card title="Chat Completions" icon="message" href="/api-reference/chat/completions">
    Full API reference for chat completions.
  </Card>

  <Card title="Embeddings" icon="magnifying-glass" href="/api-reference/embeddings/create-embeddings">
    Create embeddings for semantic search.
  </Card>

  <Card title="Text-to-Speech" icon="volume-high" href="/api-reference/audio/speech">
    Generate speech from text.
  </Card>
</CardGroup>