Qwable-v1 brings Claude-style coding agent behavior to open weights
Qwable-v1 is an that transfers Claude Fable-5 behavior into . Claude Fable-5 was an Anthropic preview model that was public for about four days, from June 9 to June 12, 2026, before being stopped worldwide under U.S. export-control rules.
Fable-5 was described as Anthropic’s strongest model at release, with 80.3% on and a price of $50 per million . Qwable-v1 was trained on 4,659 clear traces where the reasoning text was still available. The training took about 14 hours on one H200 .
With an agent , the model can produce XML tool calls in a Claude-like format, including tool names such as str_replace_editor. The model, GGUF files, and dataset are public on Hugging Face under the upstream AGPL-3.0 license.
Key points
- Qwable-v1 is an open-weights model distilled from Claude Fable-5 traces.
- Fable-5 was described with 80.3% on and a cost of $50 per million .
- The training used 4,659 traces where reasoning text was still visible.
- The model can emit Claude-like XML tool calls, including str_replace_editor.
- The model, GGUF files, and dataset are available on Hugging Face.