Subquadratic LLM Leaderboard
Indexed results from Subquadratic LLM Leaderboard. Each ranking retains its own metric and evaluation conditions.
Winogrande
subquadraticLLM · Subquadratic LLM Leaderboard · Winogrande · percent · Higher is better
- google/recurrentgemma-2bgoogle68.51%
- RWKV/rwkv-raven-14brwkv66.69%
- RWKV/rwkv-4-14b-pilerwkv65.43%
- google/recurrentgemma-2b-itgoogle64.17%
- RWKV/rwkv-raven-7brwkv62.9%
- RWKV/rwkv-4-7b-pilerwkv62.35%
- devingulliver/llama-pile-350bdevingulliver58.48%
- RWKV/rwkv-4-3b-pilerwkv58.33%
- RWKV/rwkv-raven-3brwkv57.46%
- RWKV/rwkv-raven-1b5rwkv53.91%
- RWKV/rwkv-4-1b5-pilerwkv53.83%
HellaSwag
subquadraticLLM · Subquadratic LLM Leaderboard · HellaSwag · percent · Higher is better
- RWKV/rwkv-raven-14brwkv71.25%
- RWKV/rwkv-4-14b-pilerwkv71.07%
- RWKV/rwkv-raven-7brwkv66.48%
- RWKV/rwkv-4-7b-pilerwkv66.31%
- RWKV/rwkv-raven-3brwkv59.78%
- RWKV/rwkv-4-3b-pilerwkv59.66%
- google/recurrentgemma-2bgoogle56.89%
- devingulliver/llama-pile-350bdevingulliver56.6%
- google/recurrentgemma-2b-itgoogle56.26%
- RWKV/rwkv-raven-1b5rwkv52.6%
- RWKV/rwkv-4-1b5-pilerwkv52.25%
ARC
subquadraticLLM · Subquadratic LLM Leaderboard · ARC · percent · Higher is better
- RWKV/rwkv-raven-14brwkv44.62%
- RWKV/rwkv-4-14b-pilerwkv44.45%
- RWKV/rwkv-4-7b-pilerwkv39.68%
- RWKV/rwkv-raven-7brwkv39.42%
- RWKV/rwkv-raven-3brwkv36.69%
- RWKV/rwkv-4-3b-pilerwkv36.01%
- devingulliver/llama-pile-350bdevingulliver33.19%
- RWKV/rwkv-4-1b5-pilerwkv31.83%
- RWKV/rwkv-raven-1b5rwkv31.83%
- google/recurrentgemma-2bgoogle31.4%
- google/recurrentgemma-2b-itgoogle30.97%
GSM8K
subquadraticLLM · Subquadratic LLM Leaderboard · GSM8K · percent · Higher is better
- google/recurrentgemma-2bgoogle16.15%
- google/recurrentgemma-2b-itgoogle10.08%
- RWKV/rwkv-raven-14brwkv2.12%
- devingulliver/llama-pile-350bdevingulliver0.76%
- RWKV/rwkv-4-7b-pilerwkv0.76%
- RWKV/rwkv-4-3b-pilerwkv0.68%
- RWKV/rwkv-raven-3brwkv0.45%
- RWKV/rwkv-4-14b-pilerwkv0.38%
- RWKV/rwkv-raven-7brwkv0.3%
- RWKV/rwkv-4-1b5-pilerwkv0%
- RWKV/rwkv-raven-1b5rwkv0%
MMLU
subquadraticLLM · Subquadratic LLM Leaderboard · MMLU · percent · Higher is better
- google/recurrentgemma-2b-itgoogle40.87%
- google/recurrentgemma-2bgoogle34.61%
- RWKV/rwkv-4-14b-pilerwkv26.12%
- RWKV/rwkv-raven-1b5rwkv25.96%
- RWKV/rwkv-raven-14brwkv25.92%
- RWKV/rwkv-4-1b5-pilerwkv25.77%
- RWKV/rwkv-4-7b-pilerwkv24.96%
- RWKV/rwkv-raven-3brwkv24.87%
- RWKV/rwkv-4-3b-pilerwkv24.67%
- devingulliver/llama-pile-350bdevingulliver24.66%
- RWKV/rwkv-raven-7brwkv23.64%
Average
subquadraticLLM · Subquadratic LLM Leaderboard · Average · percent · Higher is better
- RWKV/rwkv-raven-14brwkv42.09%
- google/recurrentgemma-2b-itgoogle40.86%
- google/recurrentgemma-2bgoogle40.44%
- RWKV/rwkv-4-14b-pilerwkv39.92%
- RWKV/rwkv-raven-7brwkv38.55%
- RWKV/rwkv-4-7b-pilerwkv37.95%
- RWKV/rwkv-raven-3brwkv35.81%
- RWKV/rwkv-4-3b-pilerwkv35.25%
- devingulliver/llama-pile-350bdevingulliver35%
- RWKV/rwkv-raven-1b5rwkv33.56%
- RWKV/rwkv-4-1b5-pilerwkv33.25%
TruthfulQA
subquadraticLLM · Subquadratic LLM Leaderboard · TruthfulQA · percent · Higher is better
- google/recurrentgemma-2b-itgoogle42.81%
- RWKV/rwkv-raven-14brwkv41.93%
- RWKV/rwkv-raven-7brwkv38.56%
- RWKV/rwkv-raven-1b5rwkv37.09%
- devingulliver/llama-pile-350bdevingulliver36.28%
- RWKV/rwkv-4-1b5-pilerwkv35.8%
- RWKV/rwkv-raven-3brwkv35.6%
- google/recurrentgemma-2bgoogle35.1%
- RWKV/rwkv-4-7b-pilerwkv33.65%
- RWKV/rwkv-4-3b-pilerwkv32.14%
- RWKV/rwkv-4-14b-pilerwkv32.04%
Models
LoadingBenchmarks
Explore every benchmark
LoadingThese charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.