Mixedbread推出了专门化搜索智能体Toast 1，其搜索质量与Claude Opus 5和GPT-5.6 Sol相当或更优[1]。该产品通过将搜索任务与推理任务分离来降低成本和延迟，可独立运行或作为子智能体集成到现有系统中[1]。
在性能对比上，Toast 1实现了显著优势[1]。在OfficeQA Pro V2测试中，GPT-5.6 Sol配合Toast 1达到70%正确率，每任务成本约$1.15，相比Claude Fable 5的60%正确率和$4/任务成本有明显提升[1]。在法律知识基准测试中，使用Toast 1将代币使用量从80.6M降至23M，成本降低超60%[1]。标准Toast 1运行成本约$0.023/查询，中位延迟8秒[1]。
相比其他模型方案，Toast 1的成本降低至原来的1/10，速度提升12倍[1]。API定价方面，输入代币为$0.30/百万，缓存输入代币$0.036/百万，输出代币$0.72/百万[1]。该产品现已通过Mixedbread API获取，推出期享受特价[1]。
Mixedbread has launched Toast 1, a specialized search agent designed to deliver search quality comparable to or exceeding Claude Opus 5 and GPT-5.6 Sol while reducing costs by 90 percent and improving speed by 12 times.[1] The agent can operate independently or be integrated as a sub-agent within existing systems, with the architecture separating search tasks from reasoning tasks to minimize both cost and latency.[1]
Performance benchmarks demonstrate Toast 1's efficiency across multiple domains.[1] In the OfficeQA Pro V2 evaluation, GPT-5.6 Sol paired with Toast 1 achieved a 70 percent accuracy rate at approximately $1.15 per task, outperforming Claude Fable 5, which scored 60 percent accuracy at $4 per task.[1] On legal knowledge benchmarks, deploying Toast 1 reduced token consumption from 80.6 million to 23 million, cutting costs by over 60 percent.[1] Standard Toast 1 queries cost approximately $0.023 each with a median latency of 8 seconds.[1]
The Toast 1 API is priced at $0.30 per million input tokens, $0.036 per million cached input tokens, and $0.72 per million output tokens, with introductory pricing currently available through the Mixedbread API.[1]