Best custodial-weights alternatives

Qwen2.5-Coder-7B-Instruct is a 7B parameter instruction-tuned language model optimized for code-related tasks such as code generation, reasoning, and bug fixing. Based on the Qwen2.5 architecture, it incorporates enhancements like RoPE,...

Gemma 2 9B by Google is an advanced, open-source language model that sets a new standard for efficiency and performance in its size class. Designed for a wide variety of...

Amazon Nova Micro is the fastest and most cost-effective text-only model in the Nova family, optimized for speed and low latency. Ideal for customer service, summarization, and translation at scale.

Amazon Nova Lite is a very low-cost multimodal model that can process image, video, and text inputs. Fast and accurate for a wide range of tasks requiring visual and language understanding.

Amazon Titan Text Express is a generative LLM for summarization, text generation, classification, open-ended Q&A, and information extraction. Optimized for enterprise workloads via AWS Bedrock.

AI21 Jamba 1.6 Mini is a lightweight Mamba-Transformer hybrid optimized for cost-effective, high-throughput inference with an impressive 256K context window. An excellent choice for document-heavy workloads on a budget.

LongCat-Flash-Chat is a large-scale Mixture-of-Experts (MoE) model with 560B total parameters, of which 18.6B–31.3B (≈27B on average) are dynamically activated per input. It introduces a shortcut-connected MoE design to reduce...

Qwen2.5-VL-32B is a multimodal vision-language model fine-tuned through reinforcement learning for enhanced mathematical reasoning, structured outputs, and visual problem-solving capabilities. It excels at visual analysis tasks, including object recognition, textual...