#2064 Community benchmark: RTX PRO 1000 Blackwell 8 GB, Windows, Coder IQ1_M, Strata 0.1.43

Descripción

## Summary
Community speed benchmark on an HP ZBook X G1i 16 with an NVIDIA RTX PRO 1000 Blackwell Generation Laptop GPU (8 GiB VRAM), Core Ultra 7 255H, 63.4 GiB RAM, and Windows 11. The run used Strata/engine 0.1.43, Qwen3.8-Flash-Next Coder IQ1_M, and a 65,536-token context.

## What changed
- Added the generated benchmark README and per-request `results.json` under `bench/results/2026-10-12-community-rtx-pro-1000-blackwell-generat-qwen3-8-flash-next-coder-iq1-m/`.
- Fresh-prompt prefill medians: 345.0 tok/s at 4K, 335.0 tok/s at 16K, and 314.4 tok/s at 32K (three runs each).
- Short-chat decode median: 23.5 tok/s (nine requests). Synthetic agent follow-up decode median: 21.5 tok/s; conversation cache hit: 68.7% (three sessions).

## Extra notes
- Command: `python tools/strata_bench.py --long` against an already running server. The benchmark did not generate an engine startup log.
- This is a speed report only. It does not test coding correctness or quality and is not a controlled comparison with another Strata version. Background load, power limits, and peak memory use were not recorded.
- No engine or benchmark code changes are included.