Google: Gemma 4 31B (free)

Gemma 4 31B is a Google model available here at no cost. It accepts text, images, and video as inputs, giving it multimodal range that text-only free alternatives lack. Its context window reaches 262,144 tokens, and it supports both tool use and reasoning, making it viable for agentic workflows and multi-step tasks. Outputs are capped at 32,768 tokens per completion. Structured output support is unconfirmed based on available data. The free pricing makes Gemma 4 31B worth shortlisting for developers prototyping on a budget or teams evaluating multimodal pipelines before committing to paid options. The significant caveat is benchmark coverage: no independent scores are available yet, so there is no external data to confirm how it ranks against comparable models on coding, reasoning, or instruction-following tasks. Users should treat performance as unproven and run their own task-specific tests before relying on it in production.