One-line: ⚠ no published benchmarks — agent-loop suitability unknown. Official BlinkDL RWKV-7 “Goose” G1j weight family (1.5B/2.9B/7.2B/13.3B), dropped 2026-09-02; Apache-2.0, attention-free constant-state RNN ideally suited to local deployment.
Coding benchmarks
Benchmark
Score
Source
HumanEval
—
none published
HumanEval+
—
none published
MBPP
—
none published
LiveCodeBench
—
none published
Aider polyglot
—
none published
SWE-bench Verified
—
none published
SWE-bench Multilingual
—
none published
Agentic / tool-use benchmarks
Benchmark
Score
Source
BFCL
—
none published
τ-Bench
—
none published
ToolACE
—
none published
GAIA
—
none published
Reasoning (context only)
Benchmark
Score
Source
MMLU / MMLU-Pro
—
none published
GPQA Diamond
—
none published
MATH
—
none published
AIME 2025
—
none published
Efficiency
Active params: total = all (dense/recurrent; no MoE). Sizes 1.5B / 2.9B / 7.2B / 13.3B
Context length: 16384 base (state-extrapolatable to much longer — .pth filename *ctx16384)
Successor to the G1i family (the prior run’s single qualifying model). RWKV-7 “Goose” is an attention-free recurrent architecture with constant-size recurrent state and constant inference work per token — training stays parallelizable but inference cost does not grow with context, which makes the largest 13.3B variant unusually cheap to run long-context on modest hardware. Ships a chat template and was pretrained on web, code, synthetic, instruction, chat, and reasoning data, so it is chat/tool-template-capable despite being a base checkpoint. Zero published benchmarks — no SWE/HumanEval/LiveCodeBench/Aider/BFCL numbers, so agent-loop quality is unverified. Uses a custom architecture (model_type: rwkv7, Rwkv7ForCausalLM) — stock transformers won’t load it without the repo’s custom model code (transformers 5.15+); llama.cpp GGUF support typically lags behind. Surfaced via per-org author= sweep + one-repo-many-version sibling check (raw weights at BlinkDL/rwkv7-g1, dated 20260831); invisible to the createdAt/downloads/lastModified streams because it dropped today at 0 downloads.