Gemma 4 26B-A4B on RTX 5090 workstation (1x 32 GB)

1. Decode: reading 3.8B active parameters at 5.69 bits/weight takes 1.90 ms at 1792 GB/s x 79% achieved efficiency.
2. MoE routing penalty of 15% applied: expert gathers are less bandwidth-efficient than a dense sweep.
3. Per-token overhead of 0.8 ms (kernel launches, attention bookkeeping, sampling) is significant here — this model is not purely bandwidth-bound on this hardware.
4. Prefill: 210 TFLOPS (FP16) x 2 for native FP8 tensor cores x 0.71 calibrated against measured prefill on this platform x 32% assumed model-FLOPs utilisation, divided by 2 x 9.9B parameters per token.
5. Prefill uses 9.9B effective parameters, not the 3.8B active in decode: a batch of hundreds of tokens routes across most of the expert pool.

• MoE decode depends on how well the engine batches expert gathers; real results vary more than for dense models.
• With few active parameters on fast memory, fixed per-token overhead rather than bandwidth sets the ceiling. Real engines vary widely in how well they hide it.
• Prefill throughput is highly engine-dependent. Flash attention, batch size and quantized KV all move this number substantially, and for sparse mixture-of-experts models it is the least reliable figure we produce.
• This is a calculated estimate, not a measurement. It assumes a single request, a short prompt, no speculative decoding and a warm model already resident in memory.

Estimator version estimator_v1. Stored with every estimated row so old estimates can be regenerated when the model improves.