15/4 Weekly Update: HumanX and the Gemma 4 release

Huge week for Doubleword at HumanX - one clear takeaway: more teams are moving non-real-time workloads off expensive realtime APIs.

New: Doubleword CLI

We’ve also shipped the dw CLI - making it much easier to run and scale async workloads, especially for coding agents.

You can prepare data, run batches, stream results, and track cost - all directly from your terminal (check it out below!)

Docs: https://docs.doubleword.ai/inference-api/dw-cli

New model: Gemma 4 (31B)

We’ve also added google/gemma-4-31B-it to Doubleword. It’s Google DeepMind’s most capable open model, built for advanced reasoning, coding, and multimodal tasks.

It’s now live at $0.07 / $0.20 per 1M tokens (batch) which is ~90% cheaper than comparable models like Claude Haiku 4.5 and GPT-5-mini.

If you’ve previously run batches with Qwen3.5-35B, this sits in a similar intelligence class - worth swapping the model and running the same job side-by-side to compare performance on your workload.