Fortytwo launches Engine API, cutting tokens by 24.7% on coding tasks
Fortytwo announced the Fortytwo Engine, an inference API that coordinates multiple AI models to steer reasoning and cut wasted generation. In the company’s DeepSWE benchmark of 450 software‑engineering tasks, the Engine ran the Qwen 3.8 27B model with 24.7% fewer tokens and a higher success score, reporting 5.6 percentage‑points improvement over the base model.
Key points
- Fortytwo Engine API uses Qwen 3.8 27B and reduces token usage by 24.7% on 450 DeepSWE coding runs
- The Engine achieved a score of 47.8 versus 42.2 for the base model, a 5.6‑point gain
- API is compatible with OpenAI Chat Completions endpoints, supporting streaming, function tools and parallel calls
The service mimics OpenAI’s Chat Completions API, letting developers keep existing SDKs while adding features such as streaming, function‑tool calls and parallel tool execution. Token accounting includes all participating models, and the API currently supports chat‑completion and model‑listing endpoints but not embeddings, audio or assistant calls. Fortytwo positions the Engine as a way to lower per‑task costs for agents that perform many reasoning steps.
Fortytwo, founded in 2024 with a 15‑engineer team, previously raised $2.3 million in pre‑seed funding and runs a separate on‑premise offering called Harbor. The new Engine is the first step toward broader collective‑AI services that aim to improve both answer quality and token efficiency for developers building AI agents.
Fortytwo Launches Engine API, Reporting 24.7% Fewer Tokens on Coding Tasks
Unite.AI · 15 October 2026
Loading the full article…
This text was published by Unite.AI and written by Antoine Tardif, CEO & Founder of Unite.AI. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- AWS outlines Trust in Depth for safer AI agent autonomy · 1 src
- Microsoft readies Windows for Nvidia RTX Spark laptops with agent start-up · 1 src
- Meta's Muse AI permits explicit adult conversations, unlike other Meta apps · 3 src
- Proof introduces x401 protocol to verify AI agent transactions · 1 src
- Meta rebrands Llama Stack to OGX as open-source agentic API server · 1 src
Comments
via GitHub Discussions