We're excited to announce GLM-5 on Together AI 🚀AI natives can now deploy Z.ai's latest reasoning and coding model on Together AI, and benefit from reliable inference for production-scale reasoning, coding, and agent workflows.Highlights:→ Best-in-class open-source—77.8% SWE-Bench Verified, 50.4% HLE w/ tools, closing gap with frontier models→ MoE efficiency—40B activated params with DeepSeek Sparse Attention reducing deployment costs→ #1 long-horizon agentic—$4,432 on Vending Bench 2, leading ...

Together Ai

We're excited to announce MiniMax M2.5 on Together AI 🚀AI natives can now deploy MiniMax's latest reasoning and coding model on Together AI, and benefit from reliable inference for production-scale agentic workflows.Highlights:→ Architect-level planning — Spec-writing with feature decomposition before coding, spanning full development lifecycle→ SOTA agentic coding — 80.2% SWE-Bench Verified, 37% faster than M2.1, matching Opus 4.6 speed→ Office deliverables — Word, PowerPoint, Excel outputs tr ...

Together Ai

Introducing Kimi K2.5 from Moonshot AI, a 1T parameter native multimodal thinking agent with Agent Swarm orchestration and vision-grounded coding. AI natives can now use Kimi K2.5 on Together AI and benefit from reliable inference for production-scale multimodal agentic workflows.Highlights:👉  Code with vision—turn chats, images & videos into aesthetic websites with expressive motion👉  Agent Swarm (beta)—orchestrate 100 parallel sub-agents, 4.5× faster than single-agent execution👉  Leading ag ...

Together Ai