# Together AI — CoderForge-Preview: SOTA open dataset for training efficient coding agents

- Company: Together AI (together.ai)
- Announced: 2026-02-25T00:00:00+00:00
- Category: research-paper
- Subject: Inference platform
- Models affected: Qwen-3 32B, Qwen3-Coder-480B
- Context window: 128K tokens
- Source: https://www.together.ai/blog/coderforge-preview
- Record: https://forck.live/items/2377-coderforge-preview-sota-open-dataset-for-training-efficient-coding-agents

Together AI releases CoderForge-Preview, the largest open dataset of coding agent trajectories (258k trajectories, 128K context length). Fine-tuning Qwen-3 32B on this dataset achieves 59.4% pass@1 on SWE-Bench Verified, a 23.0% improvement over the base model and ranking #1 among open-data models in the ≤32B parameter range.

## Evidence

Verbatim from https://www.together.ai/blog/coderforge-preview:

> We release CoderForge-Preview - the largest open test-verified coding agent dataset. By leveraging it to fine-tune Qwen-3 32B, we boost SWE-Bench Verified performance 23.0% above the base model reaching 59.4%, ranking #1 among open-data models in the ≤32B parameter range.

---

Record: https://forck.live/items/2377-coderforge-preview-sota-open-dataset-for-training-efficient-coding-agents
Catalogue: https://forck.live/llms.txt
Feed: https://forck.live/feed.md