PRISM ML
Inkling 2 Mini Launches with Looped Transformer Architecture for Local Laptop Deployment
Prism ML has released Inkling 2 Mini, a looped transformer model featuring 26.4B total parameters optimized for high-performance reasoning in a 5.4 GB footprint.
Prism ML introduced Inkling 2 Mini, an efficient model engineered to execute complex reasoning tasks locally on consumer hardware. The model implements a looped transformer design, executing middle layers multiple times per token. This mechanism enables prolonged internal reasoning without inflating external chain-of-thought token counts.
Boasting 26.4B total parameters with 3.4B active and 5.4B weight activations per token, Inkling 2 Mini outperforms comparable sub-100B open-weights models on rigorous benchmarks like DeepSWE v1.1 and Vals Index v2.2. Utilizing advanced quantization compilers, the model fits neatly into low-memory environments while retaining exceptional execution speed.