RWKV-7 Goose Models by RWKV Project
Details
About
RWKV-7 (codenamed Goose) is the seventh-generation RWKV architecture, introduced in the paper 'RWKV-7 Goose with Expressive Dynamic State Evolution' (March 2025, arXiv 2503.14456). Goose introduces the Generalized Delta Rule and expressive dynamic state evolution, comprehensively outperforming RWKV-6 and competitive with Transformer models of equivalent scale across reasoning, coding, and mathematical benchmarks. World3 training corpus spans 3.1 trillion tokens and 100+ languages. Available in four sizes: 0.1B, 0.4B, 1.5B, and 2.9B. All models are Apache 2.0 licensed and available via HuggingFace Transformers. RWKV-8 (Heron) is in early experimental development as of May 2025.
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
Use when the workload needs Infinity context and 2.9B parameters.
Use when the workload needs Infinity context and 1.5B parameters.
Use when the workload needs Infinity context and 450M parameters.
Use when the workload needs Infinity context and 190M parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| RWKV-7 Goose 2.9B | Use when the workload needs Infinity context and 2.9B parameters. | 2025-03 | Infinity context2.9B parameters | Current |
| RWKV-7 Goose 1.5B | Use when the workload needs Infinity context and 1.5B parameters. | 2025-03 | Infinity context1.5B parameters | Current |
| RWKV-7 Goose 0.4B | Use when the workload needs Infinity context and 450M parameters. | 2025-03 | Infinity context450M parameters | Current |
| RWKV-7 Goose 0.1B | Use when the workload needs Infinity context and 190M parameters. | 2025-03 | Infinity context190M parameters | Current |
Release Timeline
1 release groupSpecifications(4 models)
| Model | Released | Context | Parameters |
|---|---|---|---|
| RWKV-7 Goose 2.9B | 2025-03 | Infinite | 2.9B |
| RWKV-7 Goose 1.5B | 2025-03 | Infinite | 1.5B |
| RWKV-7 Goose 0.4B | 2025-03 | Infinite | 450M |
| RWKV-7 Goose 0.1B | 2025-03 | Infinite | 190M |
Frequently Asked Questions
- What is RWKV-7 Goose used for?
- RWKV-7 Goose is used for coding and math-heavy prompts. The family description and listed model capabilities point to those workloads as the best fit.
- How does RWKV-7 Goose compare to RWKV-6 Finch?
- RWKV-7 Goose by RWKV Project is strongest where you need coding, while RWKV-6 Finch by RWKV Project is the closest related family to check for coding. RWKV-7 Goose has 4 listed variants, so compare the specs and pricing tables before choosing a production model.
- Which RWKV-7 Goose model should I use?
- If price is the main constraint, use the pricing table first because RWKV-7 Goose does not have complete provider pricing in the local data. For the most capable/latest local choice, evaluate RWKV-7 Goose 2.9B.