#2087 Community benchmark: RTX 3090 | Qwen3.8-Flash-Next-UD-Q4_K_XL | Strata 0.1.43

说明

This PR adds a community benchmark for **Strata 0.1.43**, measured on October 12, 2026.

**Hardware**
- NVIDIA GeForce RTX 3090 — 24 GiB VRAM
- AMD Ryzen 9 5950X — 16 cores / 32 threads
- 64 GB RAM
- Linux / CUDA 13.0

**Model**
- Qwen3.8 Flash Next
- Quantization: Unsloth UD-Q4_K_XL
- Context: 65,536 tokens

**Benchmark results (median)**
- Short-chat decode: **33.1 tokens/s**
- 4K prompt prefill: **180.5 tokens/s**
- 16K prompt prefill: **382.9 tokens/s**
- Agent conversation cache reuse: **68.8%**

The contribution includes the three generated benchmark files:
- `README.md`
- `results.json`
- `engine-log-excerpt.txt`

The benchmark measures performance only; model accuracy and answer correctness were not evaluated.

No changes to Strata's source code.