AI Act Model Cards

Transparency documentation for every AI model served through Nextbit's API, published in accordance with the EU AI Act.

Introduction

Regulation (EU) 2024/1689 (the "AI Act") requires providers of general-purpose AI models to publish certain information about each model's identification, architecture, license, training data, and systemic risk profile, under Article 53(1)(c) and Article 53(1)(d).

Nextbit256 S.L. serves the models listed below via its inference API but did not develop or train any of them. Each model card below is a summary of the publicly available documentation published by the original model developer, together with Nextbit's own license verification and systemic risk assessment. Where information has not been published by the developer, the card says so explicitly rather than estimating or omitting it.

Model cards are transcribed faithfully from their source documentation and are not marketing material.


General-Purpose Models

Models intended for general natural language processing, reasoning, and instruction-following tasks.

ModelDeveloped by
ALIA-40BBarcelona Supercomputing Center (BSC)
DeepSeek-R1-Distill-Qwen-32BDeepSeek AI (fine-tune), Qwen Team, Alibaba Cloud (base)
GPT-OSS-20BOpenAI
Gemma 2 27B ITGoogle DeepMind
Gemma 4 26B A4BGoogle DeepMind
Ministral 3B 2512Mistral AI
Ministral 8B 2512Mistral AI
Ministral 14B 2512Mistral AI
Phi-4Microsoft Research
Qwen3-14BQwen Team, Alibaba Cloud
Qwen3-30B-A3BQwen Team, Alibaba Cloud
Qwen3.5-35B-A3BQwen Team, Alibaba Cloud

Creative & Roleplay Models

Fine-tuned models optimized for creative writing, storytelling, and roleplay.

ModelDeveloped by
Euryale L3.3 70B v2.3Sao10K (fine-tune), Meta AI (base)
MythoMax L2 13BGryphe (merge/fine-tune), Meta AI (base)
ReMM SLERP L2 13BUndi95 (merge), Meta AI (base)
Rocinante 12B v1.1TheDrummer / BeaverAI (fine-tune), Mistral AI (base)
UnslopNemo 12B v4.1TheDrummer (fine-tune), Mistral AI (base)

Was this page helpful?