╭────────────────────────────────────────────────────────────────────────╮
│                                                                        │
│  R I G M A R K   //   AGENT WORKLOAD RECEIPT                           │
│  BENCHMARKS LOCAL AI HOW CODING AGENTS ACTUALLY USE IT                 │
│  ▲  8/9 BASIC OUTPUT GATES PASSED — DO NOT HEADLINE                    │
│                                                                        │
├─ MODEL ────────────────────────────────────────────────────────────────┤
│                                                                        │
│  Qwen3.8-Flash-Next                                                    │
│                                                                        │
├─ SINGLE STREAM ────────────────────────────────────────────────────────┤
│                                                                        │
│  WORKLOAD         tok/s       Range tok/s   Last (s)    Checks         │
│  PROSE             94.7         69.5–97.7      111.6     ✓ 3/3         │
│  CODE              70.2         62.6–97.3      337.2     ✗ 2/3         │
│  STRUCTURED*      131.5       126.2–134.0        6.2     ✓ 3/3         │
│                                                                        │
│  * Predictable JSON ceiling; not general agent performance.            │
│  Decode medians are estimates; rates include streamed reasoning.       │
│                                                                        │
├─ PREFILL ──────────────────────────────────────────────────────────────┤
│                                                                        │
│  DEPTH       Cold tok/s      Cold TTFT (s)    Replay TTFT (s)          │
│  8K               2,546               3.22               3.15          │
│  32K              2,515              13.03              13.00          │
│  64K              2,430              26.97              26.95          │
│  Medians; replay is an immediate repeat, not a proven hit.             │
│                                                                        │
├─ CAPPED CONCURRENT GENERATION ─────────────────────────────────────────┤
│                                                                        │
│  Aggregate tok/s: C1 50.6 | C2 80.9 | C4 123.8                         │
│  C4: 0/12 normal stops; 0/12 with visible output.                      │
│  Capped throughput includes reasoning; not completed agent tasks.      │
│                                                                        │
├─ APPLIANCE ────────────────────────────────────────────────────────────┤
│                                                                        │
│  Hardware: 1x NVIDIA DGX Spark (GB10, 128 GB unified memory)           │
│  Topology: single Spark, TP1                                           │
│  Checkpoint: AxonZeta/Qwen3.8-Flash-Next-AZ-Focus-INT4 @ 5d2f89941ae2  │
│  Quantisation: INT4 AutoRound (GPTQ int4 routed experts), FP8 n-gram   │
│    table                                                               │
│  KV cache: fp8 (Mia recipe default)                                    │
│  Engine: TensorFold Zig engine (tensorfold-native) db281878 + Mia      │
│    Qwen3.8-Flash-Next single-Spark kit (commit 364f21a)                │
│                                                                        │
├─ SETTINGS ─────────────────────────────────────────────────────────────┤
│                                                                        │
│  SUITE: CUSTOM SETTINGS                                                │
│  Changed: --runs=3                                                     │
│  Changed: --decode-tokens=32768                                        │
│  REQUEST: {"chat_template_kwargs": {"reasoning_effort": "xhigh"}}      │
│  Temperature 0.0 | top_p 1.0 | seed 20260905 | protocol 1.3.0          │
│  Decode: 3 runs, 32768-token cap | Prefill: 8K/32K/64K, 3 pairs        │
│  Concurrency: C1/C2/C4, 3 rounds, 256-token cap; workload code         │
│  Cache isolation: random-salt-per-pair                                 │
│  Comparison ID: q38-int4-1spark-xhigh                                  │
│                                                                        │
├─ RECEIPT ──────────────────────────────────────────────────────────────┤
│                                                                        │
│  SOURCE     git:770c3106729d  •  clean                                 │
│  RIGMARK RELEASE: ahead of v0.0.4 (unreleased checkout)                │
│  JSON sha256:fed25946c9010cae…                                         │
│  SHARE THE CARD • LINK THE JSON RECEIPT • #RIGMARK                     │
│  github.com/alexellis/rigmark                                          │
│                                                                        │
╰────────────────────────────────────────────────────────────────────────╯