Mixedbread Launches Specialised Search Agent Toast 1

ArticleMixedbread TeamAnnouncement

Mixedbread has announced Toast 1, a specialised search and retrieval agent designed to handle query decomposition, evidence gathering, and context curation as a standalone model or subagent for frontier LLMs.

Key points

  1. Matches or exceeds search performance of Claude Opus 5 and GPT-5.6 Sol while being up to 10× cheaper and 12× faster according to Mixedbread.
  2. Takes over the complete search loop: decomposes initial queries, gathers evidence, checks sources, and returns curated context to the reasoning model.
  3. In the OfficeQA Pro V2 benchmark, GPT-5.6 Sol using Toast 1 in Codex achieved 70% correctness at approximately $1.15 per task, beating Claude Fable 5 on Databricks Genie (60% at ~$4).
  4. Reduced token consumption by 3.5× at an identical score on the Harvey LAB legal firm-knowledge benchmark, cutting costs by over 60%.
  5. Backend agnostic: works across existing retrieval setups and indexes, with tight integration into Mixedbread Search and the Chat Completions API.
  6. Launch pricing is set at $0.30 per million input tokens ($0.036 cached) and $0.72 per million output tokens.

Why it matters

Frontier models are expensive to run over multiple search iterations. Offloading query expansion and evidence retrieval to a specialised, cost-effective subagent frees up model context and compute for final reasoning and answer generation.

Analysis

Toast 1 demonstrates the architectural trend toward specialised agentic labour. Rather than using large generalist models for brute-force context retrieval, focused search agents can drastically reduce latency and token usage while improving end-to-end task accuracy.

Mentioned links

Found in

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • Link:primeintellect.ai

    Mixedbread Trains Toast 1 Deep-Search Agent Using Prime-RL

    A case study by Prime Intellect outlines how Mixedbread trained Toast 1 using Prime-RL and Verifiers v1. Fine-tuned from Qwen3.6-35B-A3B, the retrieval subagent matches frontier-level context gathering 12 times faster and at significantly lower cost.

    AI & LLMs· Resource

  • Link:Mixedbread

    Mixedbread: Integrated Retrieval Platform for AI Agents

    Mixedbread is an end-to-end data retrieval platform designed for AI agents. Rather than requiring developers to wire together standalone parsers, vector stores, and rerankers, it bundles multimodal embeddings, a native multi-vector database, and a dedicated search agent.

    AI & LLMs· Tool

  • Repository:mixedbread-ai/toast-harness

    toast-harness: Framework and API Interface for Mixedbread's toast-1 Search Agent

    Mixedbread's toast-harness is a Python repository designed for the toast-1 search agent. It allows developers to query the model via an OpenAI-compatible Completions API and provides the open-source agent harness used for training and running retrieval loops with exact token accounting.

    33starsPython

    AI & LLMs· Tool

  • Repository:xerj-org/xerj

    XERJ: Rust-based search engine for AI agents and RAG

    XERJ is a Rust-based search engine designed as an alternative to Elasticsearch and grep for AI workflows. It provides zero-config auto-indexing for code, documentation, logs, and PDFs, aiming to significantly cut token usage for AI agents handling search, RAG, and memory tasks.

    7.9kstarsRust

    AI & LLMs· Tool

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.