NVIDIA released Nemotron 3.5 Super VL on October 15, and we reached 11,840 aggregate output tokens/sec on one GB300 with 256 concurrent requests. This blog post is a walkthrough of all the tricks we used to get there. For context, this model is specifically designed to be
Learn how Jev Decision Models and Cline help build faster, cheaper AI agents with real-time decision-making, browser automation, and automated guardrails.
Explore how NVIDIA’s latest Nemotron 3.5 Super model performs with Cline’s agent harness, delivering benchmark-leading results and powering real-world AI coding workflows.
Meet Cline Desktop, an open-source app for open-weight models. Run parallel AI agents, automate recurring tasks, choose from 300+ models, and extend your workflow with plugins, MCP servers, and skills.
How Cline migrated its VS Code extension for 11 million developers to the new Cline SDK, built a safe A/B rollout, and reduced agent failures by 10x.
Benchmarking eight AI models on IMO 2026 problems in Cline, we found DeepSeek V4 Flash scored 30/42, clearing the gold medal cutoff for just $0.12. See how open-weight models compare with GPT-5.6 Sol, Claude Fable 5, Kimi K3, and more.
Build a code review agent with the Cline SDK using custom tools, guardrails, hooks, and a two-pass review workflow. Learn how to inspect PR diffs, filter false positives, audit tool usage, and post high-signal review comments to GitHub.
Learn how Cline evaluates and improves open weight coding agents using Terminal Bench, with practical heuristics for model performance, token efficiency, reasoning, providers, and eval optimization.
NVIDIA Nemotron 3.5 Lightning is now available in Cline, bringing a 30B MoE model with 3B active parameters, up to 1M context, and high throughput for agentic coding workflows.
Explore the best open weight models for coding in 2026. Compare Kimi K3, GLM 5.2, DeepSeek V4 Pro and Flash, and MiniMax M3 across benchmarks, pricing, model size, and developer use cases.
See how recursive self-improvement helped Cline optimize Kimi K3 and achieve an 88.8% SOTA score on Terminal-Bench 2.1 at a fraction of the cost.
A practical guide to the economics of self-hosting open-weight LLMs. Using Kimi K2.6 and real production traffic from Cline, this post breaks down GPU memory, inference, batching, pricing, and the point at which self-hosting can save millions.