[!WARNING] Experimental AFM artifact. This checkpoint targets AFM and AFMKit
v0.1.18-rc.3or newer. It may not work with other MLX runtimes or older AFM releases. Behavior, quality, performance, memory use, compatibility, and file layout may change.
This is an AFM format-v3 conversion of the official
zai-org/GLM-5.3-Flash
checkpoint at revision 04c4e9e95c5da8862dced7e5056455116f83a7e0. It is
converted directly from the official FP8 SafeTensors distribution; it is not a
re-quantization of a third-party 4-bit checkpoint.
The conversion contains 201 weight files totaling 181,694,111,440 bytes (approximately 169.2 GiB), excluding runtime memory and KV cache. Loading the checkpoint measured approximately 169.4 GB wired memory on a 512 GB Apple Silicon development machine. A smaller minimum memory configuration has not been established.
3mlx-affine-4num_nextn_predict_layers: 1201/2013,0590.afm-mlx-conversion.json.End-to-end vision quality is not qualified. Weight preservation alone is not an image-quality parity claim.
The structural layer 45 NextN tensors are present and converted. This is different from a general quality or speedup claim.
Qualification with AFMKit v0.1.18-rc.3 succeeded:
GLM embedded NextN layer at depth 1.PASS with
finish_reason=stop.draftedTokens=4, acceptedTokens=4,
headForwards=9, targetForwards=5, and zero rejection replays.This is runtime and integrity qualification, not a standard benchmark result. Throughput and correctness under other prompts, batch sizes, cache states, tool calls, and vision inputs remain unmeasured.
Use AFM built against AFMKit v0.1.18-rc.3 or newer:
afm mlx -m scouzi1966/GLM-5.3-Flash-AFM-MLX-4bit \
--mtp --reasoning-effort low --port 9999
The GLM-5.3 template does not expose a direct thinking off switch. Low reasoning effort reduces but does not eliminate reasoning output.
curl http://127.0.0.1:9999/v1/chat/completions \
-H 'Content-Type: application/json' \
-d '{
"model": "scouzi1966/GLM-5.3-Flash-AFM-MLX-4bit",
"messages": [{"role": "user", "content": "Return exactly PASS and nothing else."}],
"temperature": 0,
"max_tokens": 128,
"stream": true
}'
Original model by Z.AI. The original MIT license is included in LICENSE. The
upstream model card is preserved in UPSTREAM_README.md; upstream model claims
are not measurements of this conversion. AFM conversion and packaging are
provided by the AFM / AFMKit project.
model-language-layer-011-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-012-experts-down_proj.safetensors 1.4 GBmodel-language-layer-012-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-013-experts-down_proj.safetensors 1.4 GBmodel-language-layer-018-experts-down_proj.safetensors 1.4 GBmodel-language-layer-020-experts-down_proj.safetensors 1.4 GBmodel-language-layer-020-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-021-experts-down_proj.safetensors 1.4 GBmodel-language-layer-022-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-023-experts-down_proj.safetensors 1.4 GBmodel-language-layer-024-experts-down_proj.safetensors 1.4 GBmodel-language-layer-024-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-026-experts-down_proj.safetensors 1.4 GBmodel-language-layer-027-experts-down_proj.safetensors 1.4 GBmodel-language-layer-029-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-034-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-039-experts-down_proj.safetensors 1.4 GBmodel-language-layer-040-experts-down_proj.safetensors 1.4 GBmodel-language-layer-042-experts-down_proj.safetensors 1.4 GBmodel-language-layer-042-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-044-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-005-experts-down_proj.safetensors 1.4 GBmodel-language-layer-008-experts-down_proj.safetensors 1.4 GBmodel-language-layer-010-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-014-experts-down_proj.safetensors 1.4 GBmodel-language-layer-015-experts-down_proj.safetensors 1.4 GBmodel-language-layer-017-experts-down_proj.safetensors 1.4 GBmodel-language-layer-023-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-025-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-026-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-027-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-028-experts-down_proj.safetensors 1.4 GBmodel-language-layer-028-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-030-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-031-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-032-experts-down_proj.safetensors 1.4 GBmodel-language-layer-033-experts-down_proj.safetensors 1.4 GBmodel-language-layer-033-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-035-experts-down_proj.safetensors 1.4 GBmodel-language-layer-036-experts-down_proj.safetensors 1.4 GBmodel-language-layer-037-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-038-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-040-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-041-experts-down_proj.safetensors 1.4 GBmodel-language-layer-041-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-043-experts-down_proj.safetensors 1.4 GBmodel-language-layer-044-experts-down_proj.safetensors 1.4 GBmodel-language-layer-045-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-010-experts-down_proj.safetensors 1.4 GBmodel-language-layer-011-experts-down_proj.safetensors 1.4 GBmodel-language-layer-013-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-014-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-015-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-015-experts-up_proj.safetensors 1.4 GBmodel-language-layer-016-experts-down_proj.safetensors 1.4 GBmodel-language-layer-016-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-017-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-018-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-019-experts-down_proj.safetensors 1.4 GBmodel-language-layer-019-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-021-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-021-experts-up_proj.safetensors 1.4 GBmodel-language-layer-022-experts-down_proj.safetensors 1.4 GBmodel-language-layer-025-experts-down_proj.safetensors 1.4 GBmodel-language-layer-026-experts-up_proj.safetensors 1.4 GBmodel-language-layer-027-experts-up_proj.safetensors 1.4 GBmodel-language-layer-029-experts-down_proj.safetensors 1.4 GBmodel-language-layer-029-experts-up_proj.safetensors 1.4 GBmodel-language-layer-030-experts-down_proj.safetensors 1.4 GBmodel-language-layer-031-experts-down_proj.safetensors 1.4 GBmodel-language-layer-032-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-034-experts-down_proj.safetensors 1.4 GBmodel-language-layer-035-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-036-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-037-experts-down_proj.safetensors 1.4 GBmodel-language-layer-038-experts-down_proj.safetensors 1.4 GBmodel-language-layer-039-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-039-experts-up_proj.safetensors 1.4 GBmodel-language-layer-043-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-043-experts-up_proj.safetensors 1.4 GBmodel-language-layer-044-experts-up_proj.safetensors 1.4 GBmodel-language-layer-045-experts-down_proj.safetensors 1.4 GBmodel-language-layer-003-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-006-experts-down_proj.safetensors 1.4 GBmodel-language-layer-006-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-007-experts-down_proj.safetensors 1.4 GBmodel-language-layer-009-experts-down_proj.safetensors 1.4 GBmodel-language-layer-009-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-003-experts-down_proj.safetensors 1.4 GBmodel-language-layer-004-experts-down_proj.safetensors 1.4 GBmodel-language-layer-004-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-005-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-007-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-008-experts-gate_proj.safetensors 1.4 GBmodel-language-layer-010-experts-up_proj.safetensors 1.4 GBmodel-language-layer-011-experts-up_proj.safetensors 1.4 GBmodel-language-layer-013-experts-up_proj.safetensors 1.4 GBmodel-language-layer-016-experts-up_proj.safetensors 1.4 GBmodel-language-layer-019-experts-up_proj.safetensors 1.4 GBmodel-language-layer-020-experts-up_proj.safetensors 1.4 GBmodel-language-layer-022-experts-up_proj.safetensors 1.4 GBmodel-language-layer-024-experts-up_proj.safetensors 1.4 GBmodel-language-layer-030-experts-up_proj.safetensors 1.4 GBmodel-language-layer-032-experts-up_proj.safetensors 1.4 GBmodel-language-layer-036-experts-up_proj.safetensors 1.4 GBmodel-language-layer-038-experts-up_proj.safetensors 1.4 GBmodel-language-layer-041-experts-up_proj.safetensors 1.4 GBmodel-language-layer-045-experts-up_proj.safetensors 1.4 GBmodel-language-layer-012-experts-up_proj.safetensors 1.4 GBmodel-language-layer-014-experts-up_proj.safetensors 1.4 GBmodel-language-layer-017-experts-up_proj.safetensors 1.4 GBmodel-language-layer-018-experts-up_proj.safetensors 1.4 GBmodel-language-layer-023-experts-up_proj.safetensors 1.4 GBmodel-language-layer-025-experts-up_proj.safetensors 1.4 GBmodel-language-layer-028-experts-up_proj.safetensors 1.4 GBmodel-language-layer-031-experts-up_proj.safetensors 1.4 GBmodel-language-layer-033-experts-up_proj.safetensors 1.4 GBmodel-language-layer-034-experts-up_proj.safetensors 1.4 GBmodel-language-layer-035-experts-up_proj.safetensors 1.4 GBmodel-language-layer-037-experts-up_proj.safetensors 1.4 GBmodel-language-layer-040-experts-up_proj.safetensors 1.4 GBmodel-language-layer-042-experts-up_proj.safetensors 1.4 GBmodel-language-layer-003-experts-up_proj.safetensors 1.4 GBmodel-language-layer-004-experts-up_proj.safetensors 1.4 GBmodel-language-layer-005-experts-up_proj.safetensors 1.4 GBmodel-language-layer-006-experts-up_proj.safetensors 1.4 GBmodel-language-layer-007-experts-up_proj.safetensors 1.4 GBmodel-language-layer-008-experts-up_proj.safetensors 1.4 GBmodel-language-layer-009-experts-up_proj.safetensors 1.4 GBmodel-language-globals.safetensors 714 MBmodel-vision-globals.safetensors 321 MBmodel-language-layer-000.safetensors 164 MBmodel-language-layer-002.safetensors 164 MBmodel-language-layer-001.safetensors 164 MBmodel-language-layer-045.safetensors 106 MBmodel-language-layer-012.safetensors 96 MBmodel-language-layer-040.safetensors 96 MBmodel-language-layer-038.safetensors 96 MBmodel-language-layer-042.safetensors 96 MBmodel-language-layer-029.safetensors 96 MBmodel-language-layer-033.safetensors 96 MBmodel-language-layer-025.safetensors 96 MBmodel-language-layer-036.safetensors 96 MBmodel-language-layer-020.safetensors 96 MBmodel-language-layer-034.safetensors 96 MBmodel-language-layer-010.safetensors 96 MBmodel-language-layer-014.safetensors 96 MBmodel-language-layer-016.safetensors 96 MBmodel-language-layer-021.safetensors 96 MBmodel-language-layer-022.safetensors 96 MBmodel-language-layer-044.safetensors 96 MBmodel-language-layer-013.safetensors 96 MBmodel-language-layer-041.safetensors 96 MBmodel-language-layer-017.safetensors 96 MBmodel-language-layer-030.safetensors 96 MBmodel-language-layer-024.safetensors 96 MBmodel-language-layer-028.safetensors 96 MBmodel-language-layer-026.safetensors 96 MBmodel-language-layer-032.safetensors 96 MBmodel-language-layer-018.safetensors 96 MBmodel-language-layer-009.safetensors 96 MBmodel-language-layer-006.safetensors 96 MBmodel-language-layer-037.safetensors 96 MBmodel-language-layer-008.safetensors 96 MBmodel-language-layer-004.safetensors 96 MBmodel-language-layer-005.safetensors 96 MBmodel-language-layer-019.safetensors 89 MBmodel-language-layer-043.safetensors 89 MBmodel-language-layer-039.safetensors 89 MBmodel-language-layer-011.safetensors 89 MBmodel-language-layer-031.safetensors 89 MBmodel-language-layer-023.safetensors 89 MBmodel-language-layer-027.safetensors 89 MBmodel-language-layer-035.safetensors 89 MBmodel-language-layer-015.safetensors 89 MBmodel-language-layer-007.safetensors 89 MBmodel-language-layer-003.safetensors 89 MBmodel-vision-block-023.safetensors 34 MBmodel-vision-block-016.safetensors 34 MBmodel-vision-block-014.safetensors 34 MBmodel-vision-block-012.safetensors 34 MBmodel-vision-block-017.safetensors 34 MBmodel-vision-block-020.safetensors 34 MBmodel-vision-block-002.safetensors 34 MBmodel-vision-block-008.safetensors 34 MBmodel-vision-block-001.safetensors 34 MBmodel-vision-block-013.safetensors 34 MBmodel-vision-block-011.safetensors 34 MBmodel-vision-block-018.safetensors 34 MBmodel-vision-block-021.safetensors 34 MBmodel-vision-block-022.safetensors 34 MBmodel-vision-block-009.safetensors 34 MBmodel-vision-block-007.safetensors 34 MBmodel-vision-block-000.safetensors 34 MBmodel-vision-block-005.safetensors 34 MBmodel-vision-block-015.safetensors 34 MBmodel-vision-block-003.safetensors 34 MBmodel-vision-block-010.safetensors 34 MBmodel-vision-block-006.safetensors 34 MBmodel-vision-block-019.safetensors 34 MBmodel-vision-block-004.safetensors 34 MB有文件?提交种子。
提交种子zai-org/GLM-5.3-Flash