Moonshot AI releases Kimi K3.1 Pro, promises open weights by Nov. 4
For the CTO of a mid-size US bank
Beijing-based Moonshot AI introduced Kimi K3.1 Pro, the first model in its K3.1 series, and promised a public release of the full model weights by Nov. 4, 2026, according to a report by Crowdfund Insider.1
The company built the model on the earlier Kimi K3 with substantially increased post-training computation, including large-scale reinforcement learning across hundreds of domains, according to Crowdfund Insider.1 The report cites the company's own claims and links to a Moonshot blog post; the full text of that post was not among the sources reviewed.1
According to Crowdfund Insider, the model keeps a mixture-of-experts architecture with about 2.8 trillion total parameters and roughly 104 billion active per token, and adds a one-million-token context window and support for text plus image inputs; some descriptions also mention video, which the report does not confirm.1
The company reports 65.7 percent on Terminal-Bench 4.0, ahead of OpenAI's GPT-6 Astra and narrowly behind Anthropic's Claude Opus 5.5 in the company's published comparisons.1
Independent evaluations place the model fourth overall on the Artificial Analysis Intelligence Index shortly after launch, according to the report, which did not name the evaluator beyond the index.1
The company applied data-filtering measures during training to restrain certain vulnerability-research capabilities ahead of the open-weight release, the report said; it did not describe what data was filtered.1
Banks that plan to run open weights in-house can keep prompts off outside servers, but they cannot inspect the training choices, such as the vulnerability-research filtering, so model risk reviews will rest on vendor claims and their own testing. Use of the hosted API sends prompts to Moonshot.
Moonshot has promised to publish the full model weights by Nov. 4, 2026.1
Our take. Kimi K3.1 Pro scores 65.7% on a coding-agent test, just behind Claude Opus 5.5, and its weights go public by November 4. Your developers can download it that week; your policy on Chinese-origin open weights should exist first.
Action. Information security: Block or approve downloads of Kimi K3.1 Pro weights and Moonshot API calls under a written open-weight model policy.
What our records say
Anthropic sells banks Claude seats and API access directly and through cloud marketplaces.1 In 2026 it added embedded engineering with Goldman Sachs and FIS, restricted cyber-defence access to its Mythos model for a few large banks, and a services firm co-founded with Goldman Sachs.1 Goldman Sachs and JPMorganChase offer Claude to staff, and BankUnited and Flagstar run Claude in internal AI platforms.1
From our records
- Claude (Anthropic)vendor · banks using it (21) Goldman, JPMorgan, BankUnited · alternatives Microsoft Copilot, ChatGPT Enterprise, Gemini Enterprise
- ChatGPT Enterprise (OpenAI)vendor · banks using it (14) Morgan Stanley, BNY, Customers Bank · alternatives Microsoft Copilot, Claude, Gemini Enterprise
Sources
- ReportMoonshot AI unveils Kimi K3.1 Pro, challenging US models on coding Crowdfundinsider