niklaslenzClover 35B in the Gauntlet
RTX 5060 Ti 16 GB · llama.cpp (Clover build), Diffuspace Clover-35B-A3B-Preview (Q4_0 GGUF)
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) | ●✕✕✕✕ | 1 / 5 | 26.1 s | buzzer, answer forced | 4 (48.0 s) | 81.3 s |
| 2 | Painter (code, level 2) | ✕●✕●✕ | 2 / 5 | 59.3 s | buzzer, answer forced | 3 (36.0 s) | 102.5 s |
| 3 | Bench seating (puzzle, level 3) | ●●●●● | 5 / 5 | 16.6 s | no | 0 (0.0 s) | 23.8 s |
| 4 | Snake (code, level 4) | ●✕●✕● | 3 / 5 | 29.7 s | no | 2 (30.0 s) | 66.9 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.
The hardware line in the log files of this race used an internal description of the test machine. It was replaced by the plain description above; nothing else in the files was changed, and the replay still verifies.
Same setup, run again, not ranked
The game test is one attempt per setup: the first run is the ranked race above. Later runs of the same setup on the same machine are shown for the record. They have no rank and are never used instead of the first run. A repeat differs noticeably from the ranked run: one race is one sample, and the spread between runs of the same setup is real.
| Run | Finish time | Targets hit | Penalty loops | Raw files |
|---|---|---|---|---|
| Nov 10, 2026 Thinking off | 382.9 s | 7 of 20 | 13 | result.json race log model answers verify output |
Who made what
- Model and recipe (GGUF files, revision 32e22a02)Diffuspace
- Engine (llama.cpp pull request #31306, state cb09d4355)ggml-org (llama.cpp)X @ggml_orgGitHubHugging FaceWebsite
- Base model (Qwen3.6-35B-A3B)QwenHugging FaceX @Alibaba_QwenGitHubWebsite
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