Performance evidence
Power accepts performance work only when the measured system still preserves its execution contract. The published Qwen3.8 captures therefore use the real streaming API, pin artifact and binary identities, verify deterministic output, and report quality and workload throughput beside steady decode.
To verify the numbers directly, open Reproduction: run the model-free one-command verifier first, then replay the full RTX 4090 protocol against the same input identities.
Quality and speed by mode
Request-wide includes prompt processing, generation, HTTP, and request
overhead. Steady decode is a warmed-up, repetitive 1,024-token shape. A dash
means there is no defensible apples-to-apples capture for that cell.
What the pure-Q6_K 176.61 boundary means
The untouched 22,884,408,288-byte Q6_K artifact reached a 176.6109 token/s median across nine 1,024-token samples with prefix-FR8192 K7/S6. Its minimum was 173.2630 and seven samples were at least 175. The same-artifact full-vocabulary K7/S7 control reached 147.0207 token/s, so the peak profile improved median steady decode by 20.13%. Every output digest matched.
No model weight was requantized. The speedup comes from native MTP, exact target verification, an 8,192-row draft-only token-ID prefix, six resident recurrent snapshots, batched GPU greedy sampling, Flash Attention, full CUDA offload, batch 14, and host controls.
This is a high-coverage peak, not a service floor or universal default. On the one-pass 12-task calibration, full-vocabulary K7/S6 reached 47.032 token/s request-wide with 52.30% acceptance, while prefix FR reached 37.290 token/s with 24.82% acceptance. Eleven tasks per mode hit the output cap.
The previous mixed-artifact boundary remains separate. Its rollback-complete K7/S7 profile reached 175.2089 token/s steady decode and 83.228 token/s on the repeated 100-task workload. That 19,187,686,464-byte artifact uses Q4_0 main FFN tensors, a Q6_K MTP block, and a Q4_K draft head.
The current acceptance host is Windows 11 with an RTX 4090 and a 10-core,
20-thread Intel Xeon w5-2445. Its 0x55555 affinity mask is topology-specific,
not a portable product default.
Did quality fall?
The exact pure-Q6_K prefix-FR profile has not yet completed the repeated 100-task matrix. Its 12-task, 128-token calibration is too truncated to support a general intelligence claim. Exact target verification establishes fixed-prompt greedy output identity; it does not replace a representative quality evaluation.
For the previous mixed-artifact K7/S7 profile, no regression was observed on the fixed repeated sample:
- TBQ4 autoregressive scored 70/100 lenient and 64/100 strict.
- Full-vocabulary K7/S7 scored 76/100 lenient and 66/100 strict.
- All 900 requests completed without errors.
- Every prediction was stable across three repetitions.
- Proposal acceptance was 51.33%; replay and guard activation were both zero.
These are fixed-task accuracy proxies, not general intelligence or IQ scores. The paired differences did not reach conventional statistical significance, so the result does not prove that MTP improves general model intelligence.
The archived prefix-FR mode is useful negative evidence. Its steady peak was high, but mixed-workload acceptance fell to 25.55%, request-wide throughput fell to 27.951 token/s, and C-Eval acceptance was only 14.21%. Full-vocabulary K7/S7 removes that draft-coverage bottleneck.
Reproduce the boundary
The dedicated Reproduction page provides copyable commands, the fixed environment, input SHA-256 values, complete runner arguments, output files, and pass criteria.
The full replay has two levels:
- Offline verification recomputes archived report statistics, hashes, thresholds, and output identities without loading the model.
- Performance replay rebuilds the pinned CUDA profile, verifies the exact artifact, prompt, ACL, and binaries, then runs Power's streaming API workload.
The current gate fixes one warm-up, nine measured requests, 1,024 generated tokens per request, batch 14, greedy sampling, and a deterministic output SHA-256. The runner fails on a changed model identity, wrong backend, short output, non-deterministic digest, missing host control, or missed median gate.
Use the repository's checked-in guide and raw evidence:
- Complete Windows/CUDA reproduction procedure
- Benchmark record and all mode interpretations
- Untouched-Q6_K boundary and dynamic-quantization analysis
- Repeated 100-task quality protocol
- Current compact machine-readable evidence
- UD-Q8_K_XL heterogeneous-placement boundary
Treat a replay on different silicon, driver, display load, clock policy, model bytes, or prompt as a new result rather than silently combining it with this acceptance capture.