New player Maxwell launches V1 Flash: 35B/A3B MoE, 256K context, Apache 2.0
AdinaYakup · x · 2026-11-16
Newcomer Maxwell has released Maxwell V1 Flash, an Apache 2.0 MoE model with 35B total / 3B active parameters and 256K context, focused on Southeast Asian languages including Thai, Lao, Myanmar, Khmer and Shan. Its published benchmarks show it beating peer open models like Qwen3.6 35B/A3B, Gemma 4 26B/A4B and Nemotron 3.5 Lightning on reasoning (GPQA-Diamond 86.53) and agentic search tasks (HLE w/Tools 46.10, DeepSearchQA 83.30).
More from Models
- Nex-N2.6, a GLM-5-Next-based multimodal model, trends on Hugging Face — nex-agi · 2026-11-16
- FLUX-mimic-x: 7B video-action model tops RoboLab leaderboard at 51.4%, zero-shot across embodiments — robrombach · 2026-11-16
- MachinePulse open-sources Karpo-Decision-31B, tops Decision Index at 73.9 — Negative-Whereas3307 · 2026-11-16
- Qwen4-max launch page called out: no model card, no benchmarks, no pricing — petrusenko_max · 2026-11-16
- SLDBench v2: Claude Opus 5.5 beats classic fits using 4% of compute budget — jiqizhixin · 2026-11-16
- Grok Bot 1.0 lands with bots that join meetings, take initiative and self-improve — mattyp · 2026-11-16