Inkling 2 Mini tops intelligence-per-joule test: 82.3% accuracy at just 1.17 kJ per query
simonguozirui · x · 2026-11-21
JonSaadFalcon got early access to Inkling 2 Mini and ran it on an Intelligence-per-Watt benchmark on a single NVIDIA DGX Spark. At low effort in MXFP8 it hits 82.3% accuracy at 1.17 kJ per query (39W avg) — the best intelligence per joule of any model tested; medium effort reaches 87.2%. Comparison: Qwen3.6-35B-A3B scores 72.0% (0.43), Gemma 4 26B-A4B 63.2% (0.16), and dense Qwen3.5-27B just 63.8% while burning 17.63 kJ (0.04).
Related event: Inkling 2 Mini tops intelligence-per-watt benchmarks(2 posts)→
More from Infra
- 2M canary tokens expose stolen prompts and compromised AI providers — AccBalanced · 2026-11-21
- Unsloth release runs 26.4B Inkling 2 Mini on 16GB RAM, cuts MCP token use 87% — danielhanchen · 2026-11-21
- Seroter Daily Reading List #914: harnesses are cheap to swap, model behavior isn't — rseroter · 2026-11-21
- NVFP4 MoE experts force CPU fallback in llama.cpp; requantizing boosts Metal speed 5.8x — teortaxesTex · 2026-11-21
- Compute overbuild skeptic recants: efficiency gains cut AI compute cost ~4x per year — teortaxesTex · 2026-11-21
- AscendGym paper draws attention as Huawei compute platform research picks up — willknight · 2026-11-21