C.W.K.
Stream
Lesson 03 of 05 · published

RWKV — RNN-Like Inference at Transformer Quality

~9 min · frontier, rwkv, rnn

Level 0Scout
0 XP0/41 lessons0/12 achievements
0/100 XP to next level100 XP to go0% complete

The bet

RWKV ("Receptance-Weighted Key-Value") asks: can we get Transformer-quality outputs with RNN-style O(1) per-token inference? The answer, as of RWKV-7 ("Goose", March 2025), is "surprisingly close". RWKV is a different bet from Mamba — it leans even harder into the recurrent direction.

The key idea

RWKV uses time-mixing and channel-mixing operations that are reformulations of attention into a form that can be computed recurrently at inference. During training, attention-like parallel computation is possible; during inference, the same model can be evaluated step-by-step like an RNN.

Why this matters

  • Constant memory at inference. RWKV processes sequences in O(1) memory regardless of length. Ideal for streaming applications.
  • O(1) per-token inference cost. No attention to compute, no KV cache to grow. Fits naturally on resource-constrained devices.
  • Edge / streaming applications. Live transcription, on-device assistants, low-power scenarios where Transformer attention is impractical.

RWKV-7 "Goose" advances

Full-matrix dynamic state decay — earlier RWKV versions used vector decay; v7 uses matrix decay, making the state-update step more expressive. Theoretically more powerful than earlier RWKV variants while keeping the recurrent inference property.

Where RWKV is now

RWKV is genuinely competitive at small-to-medium scale (1B–14B). Frontier-scale RWKV (70B+) is still in early development as of 2026. The community is small but active, and the architecture has its own ecosystem of inference engines.

RWKV vs Mamba — different bets

Both reject quadratic attention. Mamba leans on selective state spaces with structured dynamics; RWKV leans on attention reformulations with explicit recurrent inference. They are not converging — they are exploring different parts of the post-Transformer design space.

External links

Exercise

Pick a streaming or edge-deployment use case (live transcription, on-device assistant, real-time monitoring). Read the RWKV-7 paper's claims about constant-memory inference. Estimate whether the use case actually saturates Transformer attention costs enough to justify a different architecture, or whether a small dense Transformer with quantization would be just as good in practice.

Progress

Progress is local-only — sign in to sync across devices.
Spotted a bug or have feedback on this page?Report an Issue

Comments 0

🔔 Reply notifications (sign in)
Sign inPlease sign in to comment.

No comments yet — be the first.