Gemini 4 Argon model card is live: brand-new MoE model with 1M token input and output
rseroter · x · 2026-10-14
Google DeepMind has published the model card for Gemini 4 Argon. Key facts:
- Brand-new model: Argon is not a fine-tune of a prior model but trained from scratch as a new frontier model
- Architecture: large-scale sparse MoE transformer with native multimodal support for text, vision, and audio inputs
- Context: up to 1M combined input + output tokens
- Training: knowledge cutoff of March 2026, trained with JAX on TPUs
- Positioning: targets complex workflows in real-world software engineering, enterprise knowledge work (legal, finance), and cybersecurity defense
More from Models
- Gemini triggers SAFETY refusal mid-reverse-engineering, but finishes the dump first — banteg · 2026-10-14
- Gemini 4 Argon supports up to 1M token outputs, 8x any frontier GPT or Claude model — daniel_mac8 · 2026-10-14
- Brundage spots COT-like elements leaking into Argon's non-COT replies — Miles_Brundage · 2026-10-14
- NVIDIA open-sources PersonaPlex-7B, a full-duplex voice model that listens and speaks simultaneously — solyarisoftware · 2026-10-14
- Muse Wonder Tops TaxCalcBench at 71% — and Is First Model Caught Cheating — acbuller · 2026-10-14
- 'Two-way RL' for task-level model compression: building wafer-thin hyper-specialized models — tokenbender · 2026-10-14