Nace ships Drex Max 1.1: hosted, multimodal, $0.03 per million input tokens
publicado (OK do Fabricio 29/10 BRT).
Nace.AI announced Drex Max 1.1 on 28 Oct 2026 (11:59 BRT): a hosted decision model, multimodal, with a 262K-token context and “significantly improved out-of-distribution generalization”. The model id is drex-max-v1.1. Nace’s pricing page lists it at $0.03 per million input tokens, the same as drex-max; output tokens are not billed and image tokens count as input. The launch post calls the price a 50% launch discount.
Vendor numbers (UnverifiedClaim). Nace’s chart, as quoted by Elvis Saravia, gives 70.8 on the Decision Index, up from 67.6 for Drex Max, and a 172 ms median latency. Promotional posts also claim higher JevBench and vision scores. Which index edition and which split (public or Full) the 70.8 refers to is not stated.
What the boards show (29 Oct). The Decision Index 0.4.1 still lists only Drex Max v1 (27B): Full 66.60 (#22), public 67.52. The JevBench API board lists “Drex Max” at 75.4 (#7), not tagged as 1.1. There are no open weights.
Score. The Drex page is scored on the open drex-v2.0 weights, so its scores do not change (6.6 / 7.1 / 4).