Slate V1 & hyperscaler swarms (Qwen3.7/Fara1.5/Command A+)
Key Questions
What performance does Qwen3.7-Max achieve in agentic tasks?
Qwen3.7-Max delivers up to 35 hours of autonomy and supports a 1M context window for agentic results. It has reached frontier-tier capabilities according to recent reviews.
How does Fara1.5-27B compare to other computer-use agents?
Fara1.5-27B outperforms baselines including OpenAI Operator and Gemini 2.5 on the Online-Mind2Web benchmark. It uses synthetic data pipelines developed by Microsoft Research.
What is Command A+ and its key features?
Command A+ is a 218B sparse MoE model from Cohere optimized for agentic workflows. It runs efficiently on as few as two H100 GPUs and is released as open-source.
What ongoing projects are mentioned in this highlight?
Ongoing work includes Command A+, MASFactory for multi-agent orchestration, and Gemini 3.5 Flash. These focus on advancing hyperscaler swarm capabilities.
What is MASFactory used for?
MASFactory is a graph-centric framework for orchestrating LLM-based multi-agent systems. It supports structured coordination in agent workflows.
How does Amazon Bedrock AgentCore help with context windows?
Bedrock AgentCore enables implementation of Recursive Language Models to break context window barriers. It supports code interpretation for extended agent memory.
What benchmark results are highlighted for Fara1.5?
Fara1.5 models (4B/9B/27B) show strong results on Online-Mind2Web using synthetic pipelines. They surpass several leading computer-use agents in browser tasks.
What is the status of the Slate V1 highlight?
The highlight is currently in developing status with multiple models and frameworks advancing in parallel.
Qwen3.7-Max delivers 35h autonomy/1M context agentic results; Fara1.5-27B beats baselines on Online-Mind2Web with synthetic pipelines. Ongoing: Command A+, MASFactory, Gemini 3.5 Flash.