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

# Autocomplete Role in Gobi Models

> Learn how the autocomplete role works in Gobi, which models to use, and how to customize prompt templates for inline code suggestions.

export const ModelRecommendations = ({role = "all"}) => {
  const parseMarkdownLinks = text => {
    const regex = /\[([^\]]+)\]\(([^)]+)\)/g;
    const parts = [];
    let lastIndex = 0;
    let match;
    let key = 0;
    while ((match = regex.exec(text)) !== null) {
      if (match.index > lastIndex) {
        const beforeText = text.slice(lastIndex, match.index);
        if (beforeText) {
          parts.push(<span key={key++}>{beforeText}</span>);
        }
      }
      const [, linkText, url] = match;
      parts.push(<a key={key++} href={url} target="_blank" rel="noopener noreferrer" style={{
        color: "#0066cc",
        textDecoration: "underline"
      }}>
          {linkText}
        </a>);
      lastIndex = regex.lastIndex;
    }
    if (lastIndex < text.length) {
      const remainingText = text.slice(lastIndex);
      if (remainingText) {
        parts.push(<span key={key++}>{remainingText}</span>);
      }
    }
    return parts.length > 0 ? parts : text;
  };
  const modelRecs = {
    agent_plan: {
      open: ["[Qwen3 Coder (480B)](https://hub.gourmand.dev/openrouter/qwen3-coder)", "[Qwen3 Coder (30B)](https://hub.gourmand.dev/ollama/qwen3-coder-30b)", "[Devstral (27B)](https://hub.gourmand.dev/ollama/devstral)", "[Kimi K2 (1T)](https://hub.gourmand.dev/openrouter/kimi-k2)", "[gpt-oss (120B)](https://hub.gourmand.dev/openrouter/gpt-oss-120b)", "[gpt-oss (20B)](https://hub.gourmand.dev/ollama/gpt-oss-20b)", "[GLM 4.5 (355B)](https://hub.gourmand.dev/openrouter/glm-4-5)", "[GLM 4.5 Air (106B)](https://hub.gourmand.dev/openrouter/glm-4-5-air)"],
      closed: ["[Claude Opus 4.1](https://hub.gourmand.dev/anthropic/claude-4-1-opus)", "[Claude Sonnet 4](https://hub.gourmand.dev/anthropic/claude-4-sonnet)", "[GPT-5](https://hub.gourmand.dev/openai/gpt-5)", "[Gemini 2.5 Pro](https://hub.gourmand.dev/google/gemini-2.5-pro)"],
      notes: "Closed models are slightly better than open models"
    },
    chat_edit: {
      open: ["[Qwen3 Coder (480B)](https://hub.gourmand.dev/openrouter/qwen3-coder)", "[Qwen3 Coder (30B)](https://hub.gourmand.dev/ollama/qwen3-coder-30b)", "[gpt-oss (120B)](https://hub.gourmand.dev/openrouter/gpt-oss-120b)", "[gpt-oss (20B)](https://hub.gourmand.dev/ollama/gpt-oss-20b)"],
      closed: ["[Claude Opus 4.1](https://hub.gourmand.dev/anthropic/claude-4-1-opus)", "[Claude Sonnet 4](https://hub.gourmand.dev/anthropic/claude-4-sonnet)", "[GPT-5](https://hub.gourmand.dev/openai/gpt-5)", "[Gemini 2.5 Pro](https://hub.gourmand.dev/google/gemini-2.5-pro)"],
      notes: "Closed and open models have pretty similar performance"
    },
    autocomplete: {
      open: ["[QwenCoder2.5 (1.5B)](https://hub.gourmand.dev/ollama/qwen2.5-coder-1.5b)", "[QwenCoder2.5 (7B)](https://hub.gourmand.dev/ollama/qwen2.5-coder-7b)"],
      closed: ["[Codestral](https://hub.gourmand.dev/mistral/codestral)", "[Mercury Coder](https://hub.gourmand.dev/inception/mercury-coder)"],
      notes: "Closed models are slightly better than open models"
    },
    apply: {
      open: ["N/A"],
      closed: ["[Relace Instant Apply](https://hub.gourmand.dev/relace/instant-apply)", "[Morph Fast Apply](https://hub.gourmand.dev/morphllm/morph-v2)"],
      notes: "Open models are not good enough for this model role"
    },
    embed: {
      open: ["[Nomic Embed Text](https://hub.gourmand.dev/ollama/nomic-embed-text-latest)", "Qwen3 Embedding"],
      closed: ["[Voyage Code 3](https://hub.gourmand.dev/voyageai/voyage-code-3)", "[Morph Embeddings](https://hub.gourmand.dev/morphllm/morph-embedding-v2)", "Codestral Embed"],
      notes: "Closed models are slightly better than open models"
    },
    rerank: {
      open: ["zerank-1", "zerank-1-small", "Qwen3 Reranker"],
      closed: ["[Voyage Rerank 2.5](https://hub.gourmand.dev/voyageai/rerank-2-5)", "Relace Code Rerank", "[Morph Rerank](https://hub.gourmand.dev/morphllm/morph-rerank-v2)"],
      notes: "Open models are beginning to emerge for this model role"
    },
    next_edit: {
      open: ["[Instinct](https://hub.gourmand.dev/gobi/instinct)"],
      closed: ["[Mercury Coder](https://hub.gourmand.dev/inception/mercury-coder)"],
      notes: "Closed models are better than open models"
    }
  };
  let rolesToShow = [];
  if (!role || role === "all") {
    rolesToShow = Object.keys(modelRecs);
  } else {
    const key = role.toLowerCase().replace(/\s|\//g, "_").replace(/-/g, "_");
    if (modelRecs[key]) {
      rolesToShow = [key];
    }
  }
  if (rolesToShow.length === 0) {
    return <div>No recommendations found for role: {role}</div>;
  }
  return <table style={{
    width: "100%",
    borderCollapse: "collapse",
    marginTop: "1rem"
  }}>
      <thead>
        <tr>
          <th style={{
    textAlign: "left",
    borderBottom: "1px solid #ddd",
    padding: "8px"
  }}>
            Model role
          </th>
          <th style={{
    textAlign: "left",
    borderBottom: "1px solid #ddd",
    padding: "8px"
  }}>
            Best open models
          </th>
          <th style={{
    textAlign: "left",
    borderBottom: "1px solid #ddd",
    padding: "8px"
  }}>
            Best closed models
          </th>
          <th style={{
    textAlign: "left",
    borderBottom: "1px solid #ddd",
    padding: "8px"
  }}>
            Notes
          </th>
        </tr>
      </thead>
      <tbody>
        {rolesToShow.map(roleKey => {
    const rec = modelRecs[roleKey];
    if (!rec) return null;
    return <tr key={roleKey}>
              <td style={{
      fontWeight: 600,
      verticalAlign: "top",
      padding: "8px"
    }}>
                {roleKey.replace(/_/g, " ").replace(/\b\w/g, l => l.toUpperCase())}
              </td>
              <td style={{
      padding: "8px",
      verticalAlign: "top"
    }}>
                {rec.open.map((m, i) => <div key={i} style={{
      marginBottom: "4px"
    }}>
                    {parseMarkdownLinks(m)}
                  </div>)}
              </td>
              <td style={{
      padding: "8px",
      verticalAlign: "top"
    }}>
                {rec.closed.map((m, i) => <div key={i} style={{
      marginBottom: "4px"
    }}>
                    {parseMarkdownLinks(m)}
                  </div>)}
              </td>
              <td style={{
      padding: "8px",
      verticalAlign: "top"
    }}>
                {rec.notes}
              </td>
            </tr>;
  })}
      </tbody>
    </table>;
};

An "autocomplete model" is an LLM that is trained on a special format called fill-in-the-middle (FIM). This format is designed to be given the prefix and suffix of a code file and predict what goes between. This task is very specific, which on one hand means that the models can be smaller (even a 3B parameter model can perform well). On the other hand, this means that Chat models, though larger, will often perform poorly even with extensive prompting.

In Gobi, autocomplete models are used to display inline [Autocomplete](../../ide-extensions/autocomplete/quick-start) suggestions as you type. Autocomplete models are designated by adding the `autocomplete` to the model's `roles` in `config.yaml`.

## Recommended Autocomplete models

<ModelRecommendations role="autocomplete" />

Visit the [Autocomplete Deep Dive](../deep-dives/autocomplete) for detailed setup instructions and configuration options.

## Prompt templating

You can customize the prompt template used when autocomplete happens by setting the `promptTemplates.autocomplete` property in your model configuration. Gobi uses [Handlebars syntax](https://handlebarsjs.com/guide/) for templating.

Available variables for the apply template:

* `{{{prefix}}}` - the code before your cursor
* `{{{suffix}}}` - the code after your cursor
* `{{{filename}}}` - the name of the file your cursor currently is
* `{{{reponame}}}` - the name of the folder where the codebase is
* `{{{language}}}` - the name of the programming language in full (ex. Typescript)

Example:

```yaml theme={null}
models:
  - name: My Custom Autocomplete Template
    provider: ollama
    model: qwen2.5-coder:1.5b
    promptTemplates:
      autocomplete: |
        `
        globalThis.importantFunc = importantFunc
        <|fim_prefix|>{{{prefix}}}<|fim_suffix|>{{{suffix}}}<|fim_middle|>
        `
```