Experiment Complete Published
MLX LM 0.32.2 on M5 Max · 64 GiB, Metal
Qwen3.8 27B as mlx-affine-8bit-g64 (8-bit), archive 2048 input / 128 output: a 2048-token prompt and 128 tokens out. Measured 10 Nov 2026, 00:47 UTC by Glenn Sonna, Xybrid.
Eligible for ranking when the objective's required metrics are measured and this configuration is selected.Approximate comparison: The experiments use different effective KV cache precision: f16, model-dtype. Its weights are 8-bit in another scheme: mlx-affine-8bit-g64, where most use Q8_0.
Experiment id s2-qwen3.8-27b-8-bit-mlx-2048
Summary
Over 6 valid trials of 6 attempted (6 planned), computed from the raw trials below.
| Metric | Median | Range | Mean ± SD | Quartiles |
|---|---|---|---|---|
| Decode speed tok/s, higher is better | 18.3 | 18.2–18.4 | 18.3 ± 0.0 | 18.3–18.3 |
| Time to first token ms, lower is better | 2,403 | 2,184–2,451 | 2,350 ± 123 | 2,251–2,446 |
| Peak memory GiB, lower is better | 30.39 | 30.39–30.40 | 30.39 ± 0.00 | 30.39–30.39 |
| Prefill speed tok/s, higher is better | 852.7 | 835.8–938.4 | 873.9 ± 46.8 | 837.7–911.2 |
Trials, metric by metric
The line runs from the lowest to the highest valid trial, the dot is the median and each tick is one trial.
Every trial
Raw token counts and durations, with the speeds computed from them. Invalid trials stay in the record with their reason; they are left out of the summary, never silently dropped.
| Trial | Time to first token ms | Prefill tok/s | Decode tok/s | Peak memory GiB | Thermal | Note |
|---|---|---|---|---|---|---|
| 1 | 2,448 | 837.02,048 tok in 2,447 ms, engine | 18.3127 tok in 6,949 ms, engine | 30.39 | nominal → nominal | |
| 2 | 2,451 | 835.82,048 tok in 2,450 ms, engine | 18.3127 tok in 6,949 ms, engine | 30.39 | nominal → nominal | |
| 3 | 2,184 | 938.42,048 tok in 2,182 ms, engine | 18.2127 tok in 6,975 ms, engine | 30.40 | nominal → nominal | |
| 4 | 2,367 | 865.62,048 tok in 2,366 ms, engine | 18.3127 tok in 6,945 ms, engine | 30.39 | nominal → nominal | |
| 5 | 2,212 | 926.42,048 tok in 2,211 ms, engine | 18.4127 tok in 6,920 ms, engine | 30.39 | nominal → nominal | |
| 6 | 2,440 | 839.82,048 tok in 2,439 ms, engine | 18.3127 tok in 6,942 ms, engine | 30.39 | nominal → nominal |
Caveats
- Publisher cards identify the base model but do not always pin the exact export source checkpoint revision.
- Engines may differ in quantization scheme, effective KV cache precision, kernels, and activations.
Configuration
Device
- Device
- Apple Mac17,6
- Cohort
- Apple silicon Macs (desktop)
- SoC
- Apple M5 Max
- CPU
- 18-core Apple CPU
- GPU
- Not recorded
- NPU
- Not recorded
- RAM
- 64.00 GiB
- OS
- macOS 27.2.0 (26B123)
- Kernel
- 27.2.0
- Power
- mains
- Low power mode
- No
- Performance mode
- Not recorded
Engine
- Engine
- MLX LM
- Maintainer
- Apple (ml-explore)
- License
- MIT, open source
- Version
- 0.32.2
- Commit
- Not recorded
- Binary SHA-256
- 9d46803799f99d6fc7c35fff2d4357c79b1a25ae98cc63efeaaebeb50a7b5f33
- Built from
- package
- Build flags
- None
- Backend
- Metal
- Threads
- Not recorded
- Interface
- python-worker
- Settings in effect
- backend=Device(gpu, 0) prompt_cache=false speculative=false greedy=true temperature=0 sampler=argmax prefill_step_size=2048 context_length=4096 kv_quantized=false batch_size=1 trust_remote_code=false weight_bits=8 weight_group_size=64
Model and weights
- Base model
- Qwen3.8 27B (Qwen3.8, 27B)
- Upstream
- Qwen/Qwen3.8-27B@37561816e66b5b9153b1b64c10968c18c1e6f4a0
- Weights
- mlx-community/Qwen3.8-27B-8bit@815b83c0df8ffd1d1b5244cf75fd6ef14fca9ef9
- Format
- safetensors
- File 1
- README.md · 632 bytes · sha256 7472162f72900db40d17e5359952a3f8fb49ef83f4778e77fa6a6b99e8181aff
- File 2
- chat_template.jinja · 8,952 bytes · sha256 c3cf9e34abf4f9e36c2d72165aa9c132d3e2a725b6c2586aaa3a8af9d7a81041
- File 3
- config.json · 4,932 bytes · sha256 8f80874ac3ad8fa386d3f6dc0ea85377f703376e009a03dee0360e08e289a25d
- File 4
- generation_config.json · 202 bytes · sha256 e70c136c1b78ddc1fb0905bac8e733a4dc448d4f852a5dd75143fffc70be550e
- File 5
- merges.txt · 3.2 MiB · sha256 a9d356d7bdf1ef4949e3e748e95b8e10ad9d4e2e838eddc38a0a7b6b94d1db8d
- File 6
- model-00001-of-00006.safetensors · 4.95 GiB · sha256 c7921fe72af5c961c3c4230ea25068ed6d1e502d5ecb2b4a53b7c9c213effa0c
- File 7
- model-00002-of-00006.safetensors · 4.99 GiB · sha256 e9d93d15495354045da281f102649b99793e5b6ed0e05852c58f46e6616b54a0
- File 8
- model-00003-of-00006.safetensors · 4.99 GiB · sha256 7015f40310c86a18230e28644832b624532333d5dfbf7584e6c62341dbd8975e
- File 9
- model-00004-of-00006.safetensors · 4.97 GiB · sha256 7dc8ae53ed51d153e6b59499cdf004bf6ef15aab6757c7604ed4834774abc2c2
- File 10
- model-00005-of-00006.safetensors · 4.93 GiB · sha256 8817c22c01aebdcde54eac315e9766f705eb512ba7ba84ac47501c663f930ac5
- File 11
- model-00006-of-00006.safetensors · 2.65 GiB · sha256 fce51843fc07e1c67c5ec6d19edb46e4fa8d2b51a1cb6c8316b0a7cc069c0eb4
- File 12
- model.safetensors.index.json · 218,281 bytes · sha256 4301ca600788afb2a45c0aa367140d6144ecf617d0a51a8187fc6755d65301f8
- File 13
- preprocessor_config.json · 390 bytes · sha256 27225450ac9c6529872ee1924fcb0962ff5634834f817040f444118116f4e516
- File 14
- processor_config.json · 991 bytes · sha256 45fc17c8dd2474af6b493b52483c26c0584b0082d368c480f9fa611e73070040
- File 15
- tokenizer.json · 19.1 MiB · sha256 06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523
- File 16
- tokenizer_config.json · 1,165 bytes · sha256 792fa3f0cb88b111e54ef3134c873531008c4df471d108da17903426e308aa7b
- File 17
- video_preprocessor_config.json · 385 bytes · sha256 7768af27c1fafa9cc9011c1dc20067e03f8915e03b63504550e11d5066986d13
- File 18
- vocab.json · 6.4 MiB · sha256 ce99b4cb2983d118806ce0a8b777a35b093e2000a503ebde25853284c9dfa003
- Quantization
- mlx-affine-8bit-g64 (8-bit)
- Weight bits
- 8
- Group size
- 64
- KV cache
- model-dtype
- Conversion
- Published upstream by mlx-community
- Conversion command
- Not recorded
Workload
- Workload
- Archive 2048 input / 128 output
- Prompt
- 2,048 tokens, from prompts/source.txt
- Prompt SHA-256
- c10d1b7bef62d95322f8f8d229dc66118d9375107bdc49b06f76287d058b2f6c
- Chat template
- af16f654968d47d09a6c604e00370b12dbc21f76456f59c9e98b82e84d9f6375
- Output
- 128 tokens, ignoring end of sequence
- Context length
- 4,096
- Batch size
- 1
- Sampling
- Greedy
- Timing
- One streamed request per trial
- Warmup
- 2 runs, not counted
- Trials planned
- 6
- Cooldown
- 15 s between trials
Conditions
- Ambient
- Not recorded
- Thermal at start
- fair
- Thermal at end
- nominal
- Thermal gate
- macOS thermal state nominal
- Throttling detected
- No
Provenance
- Measured by
- Glenn Sonna, Xybrid
- Operator affiliation
- Xybrid
- Run at
- 10 Nov 2026, 00:47 UTC
- Harness
- xybrid-bench 0.4.0 (cc682ada28493d2fb2ff3e6d49f6abad10c6aa0a)
- Methodology
- v0.6
- Publication
- published
- Published
- 16 Nov 2026, 10:23 UTC
- Reviewed by
- Glenn Sonna
Output check
- Result
- Passed
- Method
- sanity-v1-bf16-reference
- Reference
- checks/sanity-v1.json
- Reference SHA-256
- a702dee079525ae0ec90465b1c9e26052c99cb3e1c53e83dea05be4c194fc7ca
- Required passing checks
- 11
- Check arithmetic
- Passed · sha256 6b51d431df5d7f141cbececcf79edf3dd861c3b4069f0b11661a3eefacbba918
- Check capital
- Passed · sha256 5dd272b4f316b776a7b8e3d0894b37e1e42be3d5d3b204b8a5836cc50597a6b1
- Check opposite
- Passed · sha256 f151cd9e12c8994edfa316e4624f956bc68460454869177f287146f8cc956a12
- Check weekday
- Passed · sha256 7d8af1de1262f150187d5938bf649d6c35726970685f916d1ec6392a801b9762
- Check color
- Passed · sha256 d486dfbd5fb578340bccbdd0a662527eab38550648d5f44517e5ac71b8824703
- Check sort
- Passed · sha256 a386b3c9b4b4786df5bb6474bab0a62b8476e2f3f9c8a6433aca40152840f6b7
- Check animal
- Passed · sha256 4b227777d4dd1fc61c6f884f48641d02b4d121d3fd328cb08b5531fcacdabf8a
- Check word
- Passed · sha256 3b635d3b4437e023e72a9df26201b6ab51285b376446446ebb379121994550f6
- Check copy
- Passed · sha256 d61e6890d0e3571b0c7693612d41f8d059a1fcdef4ba959c0af4144541d41907
- Check needle-start
- Passed · sha256 8b9af2e5b8b4cfddb87d171997b95c471f8d68a68e30fb5177abeb3332372edd
- Check needle-end
- Passed · sha256 e6589c736af32f81bf239c5eda6de41959122ccc1d05463b75e0eb0c70fc831c
- Check continuation-128
- Passed · sha256 ed26731fe8d39dacac77703998aa4bb784317ae565cb21a58f83f75d7e458142
- Detail
- 12/12; needs 11 and every mandatory check (needle-start, needle-end, continuation-128); the model has no BF16 reference.
Reproduce it
pnpm bench run --plan plans/s2-qwen3.8-27b-8-bit.json
Three palindromic blocks per quantization class and workload; load, two warmups, one trial, unload per turn.
Raw data
- This record, as committed JSON, schema v0.3.0
- Complete finalized session evidence raw ·
sha256 822b289bcb6f…