December 2024 Summaries

Arcee AI has launched Arcee Orchestra, a no-code platform designed to build custom AI workflows using small language models (SLMs), which the company believes are more suited to enterprise needs than large language models (LLMs). Established in 2023 with the goal of making high-performance, domain-specific, and scalable SLMs accessible to companies across various industries, Arcee AI has seen significant market adoption of SLMs, which offer immediate value and efficiency at a lower cost compared to LLMs. The company attributes the success of SLMs to their innovative post-training pipeline and techniques like model merging and logit distillation, which enhance performance and efficiency. Arcee Orchestra enables SLMs to operate as "AI agents," allowing for the creation of agentic workflows that can handle multi-step tasks and produce actionable business outcomes. The company has grown from three founders to a team of 35 across five continents and has secured Series A funding, underscoring the market's confidence in the potential of SLMs for transforming enterprise AI. Arcee Orchestra is available in private beta, inviting businesses to explore its capabilities for improving productivity and performance through customized AI solutions.

The Arcee Model Engine, now available in public beta at models.arcee.ai, offers a hosted inference service designed to support the Arcee Orchestra system by providing access to a suite of small Language Models (SLMs). These models, including Virtuoso, Caller, Coder, Spotlight, and Maestro, are tailored for various tasks such as general-purpose processing, function-calling, programming, vision-language integration, and advanced reasoning. The service promises performance and flexibility for automating processes, driving complex reasoning, and building intelligent applications. Available under the Apache-2.0 license on HuggingFace, the models are designed for straightforward integration using OpenAI's inference format. The platform offers cost-efficient pricing, starting at $20/month, and aims to cater to a range of needs from lightweight tasks to complex analytical challenges.