What is RWKV-6 Finch used for?

RWKV-6 Finch is used for coding. The family description and listed model capabilities point to those workloads as the best fit.

How does RWKV-6 Finch compare to RWKV-7 Goose?

RWKV-6 Finch by RWKV Project is strongest where you need coding, while RWKV-7 Goose by RWKV Project is the closest related family to check for coding. RWKV-6 Finch has 4 listed variants, so compare the specs and pricing tables before choosing a production model.

Which RWKV-6 Finch model should I use?

If price is the main constraint, use the pricing table first because RWKV-6 Finch does not have complete provider pricing in the local data. For the most capable/latest local choice, evaluate RWKV-6 Finch 14B.

RWKV-6 Finch Models by RWKV Project

RWKV ProjectApache 2.0

4 models2024

About

RWKV-6 (codenamed Finch) is the sixth-generation architecture in the RWKV series, introduced in the paper 'Eagle and Finch: RWKV with Matrix-Valued States and Dynamic Recurrence' (April 2024). Finch upgrades the state representation to multi-headed matrix-valued states and introduces dynamic data-driven recurrence, achieving significantly better benchmark performance than RWKV-5 (Eagle) without sacrificing the architecture's defining O(n) training complexity and O(1) constant-memory inference. Models are available in 1.6B, 3B, 7B, and 14B parameter sizes. The 14B model (released September 2024) was created by stacking two 7B models and is the largest released RWKV-6 model. All World variants are multilingual (100+ languages). Apache 2.0 licensed.

Current Variants

Use-when guidance is derived from seed capabilities, context, release, and replacement fields.

4 in view

RWKV-6 Finch 14BCurrent

Use when the workload needs 14B parameters.

2024-0914B parameters

RWKV-6 Finch 7BCurrent

Use when the workload needs 7B parameters.

2024-047B parameters

RWKV-6 Finch 3BCurrent

Use when the workload needs 3B parameters.

2024-043B parameters

RWKV-6 Finch 1.6BCurrent

Use when the workload needs 1.6B parameters.

2024-041.6B parameters

Current RWKV-6 Finch variants with use-when guidance and lifecycle status
Model	Use when	Released	Signals	Status
RWKV-6 Finch 14B	Use when the workload needs 14B parameters.	2024-09	14B parameters	Current
RWKV-6 Finch 7B	Use when the workload needs 7B parameters.	2024-04	7B parameters	Current
RWKV-6 Finch 3B	Use when the workload needs 3B parameters.	2024-04	3B parameters	Current
RWKV-6 Finch 1.6B	Use when the workload needs 1.6B parameters.	2024-04	1.6B parameters	Current

Release Timeline

2 release groups

2024-09

1 current

RWKV-6 Finch 14B

14B parameters

Current

2024-04

3 current

RWKV-6 Finch 1.6B

1.6B parameters

Current

RWKV-6 Finch 3B

3B parameters

Current

RWKV-6 Finch 7B

7B parameters

Current

Specifications(4 models)

RWKV-6 Finch model specifications comparison
Model	Released	Context	Parameters
RWKV-6 Finch 14B	2024-09	Infinite	14B
RWKV-6 Finch 7B	2024-04	Infinite	7B
RWKV-6 Finch 3B	2024-04	Infinite	3B
RWKV-6 Finch 1.6B	2024-04	Infinite	1.6B

Frequently Asked Questions

What is RWKV-6 Finch used for?: RWKV-6 Finch is used for coding. The family description and listed model capabilities point to those workloads as the best fit.
How does RWKV-6 Finch compare to RWKV-7 Goose?: RWKV-6 Finch by RWKV Project is strongest where you need coding, while RWKV-7 Goose by RWKV Project is the closest related family to check for coding. RWKV-6 Finch has 4 listed variants, so compare the specs and pricing tables before choosing a production model.
Which RWKV-6 Finch model should I use?: If price is the main constraint, use the pricing table first because RWKV-6 Finch does not have complete provider pricing in the local data. For the most capable/latest local choice, evaluate RWKV-6 Finch 14B.