GLM-4.7-Flash · file← GLM-4.7-Flash
Q4_K_M GGUF
A smaller copy for comparison. Downloaded, never tested, then deleted. Unsloth.
- Size
- 18.3 GB
- Status
- Deleted 7 Sep 2026
- Scored runs
- 0
ResultsNever scored
This file never completed a scored run. Its record below says why.
The recordHow it was made · how it was served · what went wrong · what happened to it
Rebuilt from the lab's records by Claude, an AI model. The results above come straight from the automated runs. How these pages are made.
| File | glm-4.7-flash/GLM-4.7-Flash-Q4_K_M.gguf |
| Format | GGUF Q4_K_M, 18,312,339,808 B (HF listing) |
| Where from | downloaded 2026-08-26 from unsloth/GLM-4.7-Flash-GGUF |
| Status | deleted 2026-09-07 17:36Z |
| Hugging Face | as above |
Where it came from
- Repo / file:
unsloth/GLM-4.7-Flash-GGUF/GLM-4.7-Flash-Q4_K_M.gguf, 18,312,339,808 B (HF listing). 17.05 GiB on disk. Revision and sha256 not on record. - Command: the Q8 downloader re-made by
sedwith the file swapped, a scratch file. "downloading Q4_K_M, pid 80918" at 2026-08-26 11:36Z. - Who asked: Alan, "Try the q4 just now then we have both ends before doing the in between".
How it was served
| label | dates (UTC) | binary and version | ctx / -np / KV / -fa / --jinja / --reasoning-format | harness profile | why this label exists |
|---|---|---|---|---|---|
| glm-4.7-flash-q4km (planned, never run) | — | llama.cpp b10618 | planned -c 32768 -np 1 -cb -ngl 99 --jinja: "-np 1 DELIBERATELY, even though 18.3GB could support 4 slots", to match the Q8 arm | extraction | the Q4 end of the intact-vs-compressed comparison |
Runner a one-off script, "queued behind Q8". No run-log entry, no result file. No seats.
Decisions
- Downloaded 2026-08-26 as the Q4 end of the ladder.
- Never run. Why is not on record. That afternoon's GPU queue went to Qwen3-Next-80B instead, and the report records "A GLM-4.7-Flash Q4_K_M arm (18.3 GB) is downloaded but has not run".
- Deleted unscored 2026-09-07 17:36Z: "duplicate quants, keeping the better" and "rm glm-4.7-flash Q4_K_M (kept Q8_0)". Alan: "Go for it".
What went wrong
- Its 17 GiB beside the Q8_0 made the folder read 46.7–47 GB. That figure was then quoted as the model's size in a lab note and in our working sessions.
What happened to it
Downloaded 2026-08-26 → never benched → deleted 2026-09-07 17:36Z. No RESTORE entry. Re-downloadable from unsloth/GLM-4.7-Flash-GGUF.