ANEForge on the Hub
Collection
3 items • Updated • 1
chip string | model_identifier string | hardware_hash string | p_cores int64 | e_cores int64 | gpu_cores int64 | ram_gb int64 | macos_version string | macos_build string | aneforge_version string | power string | contributor string | peak_fp16_gemm_tflops float64 | bandwidth_gbps float64 | ridge_flop_per_byte float64 | peak_perf_per_w_gflops float64 | decode_tok_s float64 | matmul_inf_cliff string | slice_x16_cliff string | reduce_exact_sum string | timestamp_utc timestamp[s] | peak_perf_per_w_source string | cpu_levels list |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Apple M1 | MacBookAir10,1 | bb9e71be3ae6 | 4 | 4 | 7 | 8 | 15.6.1 | 24G90 | 0.4.1.dev22 | ac | Rayan-and-beyond | 4.33 | 5.4 | 802 | 622 | 278 | ~32759 | clamp @ 4094 | <= 2048 | 2026-09-12T13:27:19 | null | null |
Apple M1 | MacBookPro17,1 | dcf5b8a7dd09 | 4 | 4 | 8 | 16 | 26.6.2 | 25G83 | 0.2.1.dev42 | ac | diegobauavi | 2.7 | 0.85 | 3,171 | 484 | 276 | ~32759 | clamp @ 4094 | <= 2048 | 2026-10-02T21:38:22 | null | null |
Apple M1 Max | MacBookPro18,2 | 8dc9588b5db6 | 8 | 2 | 32 | 32 | 26.5.1 | 25F80 | 0.2.0 | ac (high-power) | sbryngelson | 4.59 | 7.44 | 616 | 897 | 244 | ~32759 | clamp @ 4094 | <= 2048 | 2026-08-05T20:14:25 | null | null |
Apple M2 Max | Mac14,6 | 8c6229361be7 | 8 | 4 | 38 | 64 | 26.7.1 | 25G241 | 0.1.dev362 | ac (high-power) | Jeff271828 | 5.38 | 8.41 | 640 | 844 | 277 | ~32759 | clamp @ 4094 | <= 2048 | 2026-10-06T05:29:29 | powermetrics | null |
Apple M2 Pro | Mac14,12 | 723396ad9091 | 6 | 4 | 16 | 32 | 26.5.2 | 25F84 | 0.2.1.dev33 | ac | axiom-of-choice | 3.36 | 0.98 | 3,429 | 744 | 276 | ~32759 | clamp @ 4094 | <= 2048 | 2026-08-05T17:34:43 | null | null |
Apple M4 | Mac16,12 | d4109cd41a75 | 4 | 6 | 8 | 16 | 26.2 | 25C56 | 0.3.1.dev7 | ac | sbryngelson | 5.44 | 7.67 | 708 | 1,112 | null | ~32759 | exact (no clamp) | <= 2048 | 2026-08-06T00:46:45 | null | null |
Apple M4 Max | Mac16,6 | c77b1414be16 | 10 | 4 | 32 | 36 | 26.7 | 25G229 | 0+unknown | ac | n0madic | 8.87 | 23.02 | 385 | 1,103 | null | ~32759 | exact (no clamp) | <= 2048 | 2026-09-16T06:08:20 | null | null |
Apple M4 Pro | Mac16,7 | 7fdf43ae2309 | 10 | 4 | 20 | 48 | 27.0.1 | 26A434 | 0.4.1.dev36 | ac (high-power) | hyfjjjj | 6.2 | 20.42 | 304 | 1,092 | 336 | ~32759 | exact (no clamp) | <= 2048 | 2026-10-02T07:43:59 | null | null |
Apple M5 | Mac17,3 | 019d3ac27339 | 4 | 6 | 8 | 16 | 26.5.2 | 25F84 | 0.1.dev277 | ac | danlee2002 | 8.71 | 20.22 | 431 | 1,184 | 228 | ~32759 | exact (no clamp) | <= 2048 | 2026-08-10T20:26:53 | null | null |
Apple M5 Pro | Mac17,8 | c38210cfc8ea | 6 | 12 | 20 | 48 | 26.5.1 | 25F80 | 0.1.4.dev32+gb8dc90fe6.d20260624 | ac (high-power) | sbryngelson | 10.04 | 24 | 418 | 918 | null | ~32759 | exact (no clamp) | <= 2048 | 2026-08-01T16:39:18 | null | null |
Apple M6 | Mac18,5 | 8701b924e26c | 2 | 4 | 12 | 16 | 27.0.1 | 26A434 | 0.4.1.dev39 | ac | 751K | 10.77 | 18.93 | 569 | null | 509 | ~32759 | exact (no clamp) | <= 2048 | 2026-10-03T15:49:20 | null | [
{
"name": "Super",
"cores": 2
},
{
"name": "Performance",
"cores": 4
},
{
"name": "Efficiency",
"cores": 6
}
] |
ANE Rooflines
Cross-Apple-Silicon performance and fp16-correctness measurements for the Apple Neural Engine (ANE), collected with ANEForge. Each row is one machine (grouped by hardware hash; identical silicon in different chassis stays distinct by model identifier).
See it charted: the ANE leaderboard ranks these machines by peak GEMM, perf-per-watt, and decode throughput.
These are community-contributed submissions mirrored from the public
bench/results/rooflines/
in the repo. The full per-size sweeps live in raw/; rooflines.json is the flattened
headline table; ROOFLINES.md is the human-readable version.
from datasets import load_dataset
ds = load_dataset("aneforge/ane-rooflines") # the flattened headline table
Columns
| column | meaning |
|---|---|
chip, model_identifier, hardware_hash |
machine identity |
p_cores, e_cores, gpu_cores, ram_gb |
CPU perf/efficiency cores, GPU cores, unified memory |
macos_version, macos_build, aneforge_version |
software the run was recorded under |
power |
ac, ac (high-power), or battery at run time |
contributor |
GitHub handle who submitted the run |
peak_fp16_gemm_tflops |
headline compute peak (measured on every machine) |
bandwidth_gbps, ridge_flop_per_byte |
streaming bandwidth and the ridge point |
peak_perf_per_w_gflops |
peak GFLOP/s per watt |
decode_tok_s |
single-stream LLM decode throughput |
matmul_inf_cliff, slice_x16_cliff, reduce_exact_sum |
fp16 correctness cliffs (magnitudes where the engine silently returns a wrong answer) |
timestamp_utc |
when the run was recorded |
Reading notes
peak_fp16_gemm_tflopsis the most robust cross-chip number. Bandwidth/ridge come from a streaming sweep and are more dispatch-overhead-sensitive on smaller/older parts, so treat them as indicative.decode_tok_scurrently reports only on A16+ (e.g. M5). The decode benchmark's 32000-vocab head exceeded the 16384 max matmul dimension on the A13-A15 families; a tiled head fixes this and the older machines re-run to populate it. Blank means the run predates the tiled head, not that the chip cannot decode.- Correctness cliffs are magnitude thresholds, independent of clock, so they are valid
even on battery.
matmul~fp16_max/2(~32752);sliceclamps|value|>4094on pre-A16 parts and is exact on A16+;reduceis bit-exact for integer sums up to 2048.
Contribute your chip
Run the suite on any Apple Silicon Mac and open a PR:
PYTHONPATH=. python3 bench/roofline_suite.py --contributor <your-gh-handle>
See the roofline drive. More chip generations sharpen the per-family map.
Cite
Bryngelson, S. H. ANEForge: Python for direct computation on the Apple Neural Engine. arXiv:2606.17090 (2026).
- Downloads last month
- 1,648