Kardashev-0.7: a trained swarm of 32 models claims frontier-level performance at 1% of inference cost
ZeroStateReflex · x · 2026-10-07
A team introduced Kardashev-0.7, billed as the first trained swarm of 32 distinct models. Using RL for Population Scaling (RLPS), the models organically develop specialization and complementary capabilities, claiming frontier-level performance at 0.007x–0.02x inference cost and 0.03x memory. The team frames it as scaling intelligence by model count toward "civilizations of models." Figures are self-reported and unverified.
Related event: Kardashev-0.7 Trains a Swarm of 32 Models via RL at Fractional Cost(3 posts)→
More from Models
- Perplexity's open-weights pplx-decider-v1.1-27b tops Hugging Face Decision Index 0.3 — AravSrinivas · 2026-10-07
- AutoAWQ Author: Reproduce Bonsai 2-Class Ternary Model for ~$43k on One B300 Node in ~4 Weeks — airesearch12 · 2026-10-07
- SuperGrok users hit Grok Bot usage limits fast, calling for a 1.5x bump — nima_owji · 2026-10-07
- NVIDIA's Nemotron Labs partners with Artificial Analysis on open-model evaluation — NVIDIAAI · 2026-10-07
- Mistral Large 4 generates a Japanese-inspired floating voxel island, sparking 'Is the EU back?' buzz — kevinkern · 2026-10-07
- Marin 535B-A23B open model training crosses halfway, Percy Liang shares learnings — ericjang11 · 2026-10-07