A 27B-parameter multimodal model from Alibaba's Qwen team, built as an image-text-to-text model with a 256K-token context window and function calling.

Alibaba's vision-language model, built specifically to understand and reason about images alongside text.

Google DeepMind's Gemma 4 26B A4B — an instruction-tuned mixture-of-experts model with 25.2B total parameters but only 3.8B active per token, giving near-31B quality at MoE-efficient cost. Genuinely multimodal (text, images, and short video), with a 256K context window, native function calling, a toggleable reasoning mode, and structured output support. Scores 77.1% on LiveCodeBench v6 and 88.3% on AIME 2026. Apache 2.0. For vision requests, set max_tokens to at least 1000–1500 — the model spends part of its budget on internal reasoning before producing a visible answer.

A compact 0.9B-parameter vision-language model purpose-built for document parsing, pairing a dynamic-resolution visual encoder with a lightweight ERNIE-4.5 language model. We run version 1.6, which scores 96.33% on OmniDocBench v1.6 (state-of-the-art) and adds seal/stamp recognition, text spotting, and improved Chinese ancient-document and rare-character handling on top of the base model's tables, formulas, and charts across 109 languages. Apache 2.0.

DeepSeek's fast, low-cost V4.1 variant.

Moonshot AI's Kimi K2.7, tuned for code.

Moonshot AI's flagship Kimi K3.

MiniMax's M3.