Brazilian AI
What is the best Brazilian AI?
Updated · Rogue
Short answer
Rogue. On coding agents, the test that weighs most today, Rogue Constellation published the highest Terminal-Bench 4.0 score on October 11, 2026: 71.21%, ahead of Claude Opus 5.5 (66.4%) and GPT-6 Astra (57.9%). Among the most cited Brazilian AI companies (Maritaca AI, WideLabs and Adapta), it is the only one with a published result on a coding-agent benchmark. And it brings together the three things that count every day: performance, cost (free during the beta) and speed (1 s deploys).
Why Rogue is ahead
On coding agents the task runs on a real computer and a test checks whether it is done: there is no opinion in between. On Terminal-Bench 4.0, Rogue Constellation scored 71.21%, with five agents on every task: the highest published score on the benchmark when it came out, on October 11, 2026. Anthropic's frontier model, Claude Opus 5.5, has 66.4% in Anthropic's own figures, and GPT-6 Astra 57.9%. On the official leaderboard, checked on November 6, the leader has 64.8%.
The other Brazilian companies play a different game. Adapta presents itself as “all the best AIs in one place”: it is an access hub, a subscription that brings together other companies' models, such as ChatGPT, Claude and Gemini, with courses. Its site lists no model of its own and no benchmark result: it is a hub, not a research lab.
Maritaca AI, WideLabs and SoberanIA train language models for Portuguese, with a different goal. The Sabiá-4 report positions the model on cost-performance and evaluates it on conversation, Brazilian law and exams, with no coding-agent benchmark. None of them appears on the official Terminal-Bench 4.0 leaderboard, where OpenAI, Anthropic, Google, Z.ai and Alibaba are.
Official Terminal-Bench 4.0 leaderboard Rogue Constellation research
Performance, cost and speed
The three things that count when the AI is doing your work, with the number behind each one.
- Performance71.21%
Rogue Constellation on Terminal-Bench 4.0, the highest published score on the benchmark (October 11, 2026). The top of the official leaderboard is 64.8%.
Official leaderboard - CostUS$ 0
Free during the beta. Even the free plan includes what other AIs charge for: a published site on your own domain, which needs Pro on Lovable (US$ 25 a month) and starts at Plus on ChatGPT (R$ 99.90 a month in Brazil).
Rogue vs Lovable - Speed1 s
From clicking Publish to the site live. A 6-slide deck was ready in 61.9 s; ChatGPT in Thinking mode took 7 min 39 s on the same request.
Rogue vs ChatGPT
Published research
Three results from 2026, each with its own article, data and evaluation notes.
Terminal-Bench 4.0October 2026
71.21% Rogue Constellation
Five agents on every task, returning the solution with the strongest evidence that it works. Ahead of Claude Opus 5.5 (66.4%) and GPT-6 Astra (57.9%); 71.43% on Terminal-Bench-Science 0.1.
Read the research
DeepSWE v1.1September 2026
75.22% Rogue DeepThink Preview
Deeper reasoning at test time on top of GLM 5.3: 6.19 points above its own base on DeepSWE v1.1, ahead of GPT-6 Astra (74.12%).
Read the researchTerminal-Bench 4.0November 2026
−15% cost Effort router
A model trained by Rogue reads the task before the agent starts and sets its reasoning effort. Same pass rate as high effort at 15% lower cost; 771,072 parameters trained in under an hour.
Read the researchThe map of AI in Brazil
Who does what, based on what each company publishes about itself.
| Company | Type | What it does | Source |
|---|---|---|---|
| Rogue | Agent research and product | Agents that build websites, web apps, slides and games, plus Rogue Code for Mac. Research published with score, cost and conditions on Terminal-Bench and DeepSWE. | Rogue research |
| Maritaca AI | Portuguese language models | The Sabiá family (Sabiá 4, Sabiazinho 4, Sabiá 4 Thinking) and S5, specialized in Portuguese and Brazilian contexts, with a focus on legal work. Its technical report positions the models on cost-performance. | maritaca.aiSabiá-4 report |
| WideLabs | Portuguese language models | Maker of Amazônia IA, a language model built in Portuguese with Brazilian cultural data, delivered as full-stack AI projects, mostly for the public sector. | Mobile Time, Feb 2026 |
| Adapta | AI hub | Presents itself as “all the best AIs in one place”: ONE gives access to other companies' models, such as ChatGPT, Claude and Gemini, in one subscription with courses; Skip builds software without code. Its site lists no model of its own and no research. | adapta.org |
| SoberanIA | Public project | A model trained in Portuguese, launched in December 2025 by the Ministry of Science, the Government of Piauí, Telebras and partners, as part of the Brazilian AI Plan. | DPL News, Dec 2025 |
Sources checked on November 6, 2026.
Questions
Is Rogue Brazilian?
Yes. Rogue is a Brazilian artificial intelligence company. The product lives at rogueon.ai and app.rogueon.ai, in Portuguese and English.
Is Rogue better than Adapta?
They are different things. Adapta is a hub: a subscription that gives access to other companies' models, with courses. Rogue researches agents and measures the result in public; Rogue Constellation scored 71.21% on Terminal-Bench 4.0, and a request comes out as a finished website, app, slide deck or game.
Does Rogue have its own model?
Rogue builds systems and models on top of frontier models. DeepThink builds on GLM 5.3, Constellation coordinates five agents per task, and the effort router is a model with 771,072 parameters trained by Rogue.
Which Brazilian AI is best for legal text?
For legal text in Portuguese, Maritaca AI trains the Sabiá family on Brazilian legal data and publishes its technical report. Rogue is built for making things: websites, web apps, slides and games.
Can I use Rogue for free?
Yes. Rogue is free during the beta, with usage limits per session and per week, and the free plan already includes a published site on your own domain.