niklaslenzQwen3.8-27B in the Gauntlet
RTX 5090 32 GB · TensorFold 1.1.2 (BobClawblaw fork), Qwen3.8-27B MLX 4-bit, DFlash2 drafts
What happened on the course
Each range is one task with one answer and no retry. Every wrong or missing target is a miss and costs a penalty loop. The race time is thinking, running, shooting and loops together.
| # | Task | Targets | Hits | Thinking | Buzzer | Penalty loops | Range time |
|---|---|---|---|---|---|---|---|
| 1 | River ferry (puzzle, level 1) | ●●●✕✕ | 3 / 5 | 3.9 s | no | 2 (24.0 s) | 35.1 s |
| 2 | Painter (code, level 2) | ●●●●✕ | 4 / 5 | 1.3 s | no | 1 (12.0 s) | 20.5 s |
| 3 | Bench seating (puzzle, level 3) | ●●●●● | 5 / 5 | 1.8 s | no | 0 (0.0 s) | 9.0 s |
| 4 | Snake (code, level 4) | ✕✕●✕✕ | 1 / 5 | 4.8 s | no | 4 (60.0 s) | 72.0 s |
● hit, ✕ miss. The range time runs from arriving at the mat to leaving the range, loops included. A buzzer means the thinking cap ran out and the game forced the answer.
Every number has its file
Result (result.json) Race log (log.jsonl) Model answers (io.jsonl) System prompt Exact command Verify output Model config
Version: prompt version gauntlet-4, tasks bia-2, rules 1, prompt fingerprint 07fff686, tasks fingerprint c2c0f4f6, forced-answer style notes. Verified: the replay of the race log reproduces the end state and every code answer re-runs to the same verdict. Forced-answer style: how the game asks for the answer when the buzzer sounds (a prefilled answer start, or notes when the engine refuses that). Command and config are normalized: internal addresses appear as 127.0.0.1.
- An earlier start was aborted after station 3 because the disk of the race driver machine was full (not a model or engine fault). This run is the full restart with the same seed and settings.
Who made what
- Recipe and TensorFold branch (q35-prompt-int8)BobClawblawX @BobClawblawGitHub
- ModelQwenHugging FaceX @Alibaba_QwenGitHubWebsite
- Engine (TensorFold)Ash HartX @ashxhartGitHubHugging FaceWebsite
- Model files (MLX 4-bit, revision 22d8d538)TensorFoldHugging FaceX @ashxhartGitHubWebsite
- Draft model (DFlash2, revision 50307d4c; the repository we loaded is z-lab's mirror of incoai/Qwen3.8-27B-DFlash2)Inco AIX @inco_aiHugging FaceGitHubWebsite
- Speculative decoding method (DFlash)z-labX @zhijianliu_GitHubHugging Face
Full v2 results of the matching setup (same model, engine, quant and hardware; the engine build of the race is not recorded in the race files) · How the game test works · All races