Hacker News
Mistral Launches Medium 3.5 (128B Dense) with Remote Coding Agents in Vibe — 77.6% SWE-Bench Verified, 497 Points on HN
Mistral AI released Mistral Medium 3.5, a dense 128B model with 256K context that unifies chat, reasoning, and coding in a single set of weights, scoring 77.6% on SWE-Bench Verified. The bigger news for builders: Vibe now supports remote coding agents — async sessions that run in the cloud, can be spawned from CLI or Le Chat, and keep running while you step away. Le Chat's new Work mode layers on a multi-step agent for research and cross-tool actions. The model's $1.50/M input pricing drew criticism on HN (497 points, 230 comments) as uncompetitive against Qwen 3.6 27B for local inference.
↳ Follow the thread