Mistral Large 4 and Llama 4 Maverick are both sparse mixture-of-experts models with a one-million-token context window on the model card, both accept images and text, and both can be downloaded and run on your own hardware. Maverick shipped on 5 April 2025 with 400 billion total parameters and 17 billion active; Mistral Large 4 previewed on 6 October 2026 and published its weights on 29 October, with 1.05 trillion total and 52 billion active, leaving nineteen months between them. At the median price across Maverick's hosting market, only about 1.3 times the cost separates them. Mistral Large 4 uses 1.05T total parameters across 128 routed experts plus one shared expert versus 402B on Maverick across 128 experts, listed as 400B by Meta; active parameters are 52B per token versus 17B, a per-token compute gap of about three to one. Mistral Large 4 exposes low, medium and high effort reasoning with high as default, while Artificial Analysis lists Maverick as having no reasoning mode. Mistral's direct price is $0.70 input and $2.10 output per million, with cached input at $0.07; Meta sells no first-party Maverick list price, hosts charge as little as $0.17 input and $0.60 output, and Artificial Analysis puts the provider median at