Gemma 4 31B

Source: https://github.com/HDZucht/Mac-Mini-M6-LLM-benchmarks Raw measured rows: https://github.com/HDZucht/Mac-Mini-M6-LLM-benchmarks/blob/3990a8e0fa5f2762911689439ce12e1c28d4c989/data/speed_2026-10-03.jsonl Hardware/method/model labels: https://github.com/HDZucht/Mac-Mini-M6-LLM-benchmarks/blob/cb99870e05aa9f5783c5c6987ddd9154834123ab/README.md Loaded context metadata: https://github.com/HDZucht/Mac-Mini-M6-LLM-benchmarks/blob/main/data/loaded_context_2026-10-03.jsonl Harness: https://github.com/HDZucht/Mac-Mini-M6-LLM-benchmarks/blob/main/bench_speed.py Original author: HDZucht / Hans-Dieter Zucht, with repository-credited automated measurements by Argus. Campaign dated October 3–4, 2026; these exact records are in the October 3 raw file. No local reproduction was performed. Headline: 7.5 tokens/s is the median of three English-prose runs (Ruhr economy). It is client-observed streaming decode throughput, not a pooled rate, overall mean, peak, native kernel result or prompt-processing rate. Five workload medians and all 15 runs are preserved separately below. Hardware: Apple Mac mini M6 (Mac18,5), 12 CPU cores (2 Super, 4 Performance, 6 Efficiency), 12 GPU cores, 32 GB unified memory, macOS 27.0.1. One machine, one model loaded alone. Model storage: ORICO 2 TB external NVMe in an ASMedia ASM246x enclosure, source-reported 40 Gb/s PCIe 4.0 x4. Apple automatic fan control. Per-run temperature, clock, power, swap, exact GPU offload and peak resident memory were not logged. Exact measured API alias: google/gemma-4-31b-qat. Display model: Google Gemma 4 31B QAT. The author names the dense Gemma 4 31B QAT model; no model-family ambiguity beyond missing exact artifact. Source Format-column label: QAT. Exact measured quantization: unknown. Measured weight format: unknown; source label is not a confirmed weight container. Application: LM Studio. Backend: unreported for this alias. Exact app/backend versions are unknown. Do not turn QAT, MLX or Splash labels into an assumed Q4 quantization. The current exact LM Studio alias links the verified QAT GGUF collection. QAT and an available Q4_0 file are not proof of measured quantization; benchmark format/backend/quant remain unassigned. Current alias corroboration (not original run metadata): https://www.lmstudio.ai/models/google/gemma-4-31b-qat Verified public related package/collection: https://huggingface.co/lmstudio-community/gemma-4-31B-it-QAT-GGUF Measured Hugging Face repository, filename, revision and hash are unrecorded. Public related-package links above prove availability only. Loaded context for this exact alias is 32,768 tokens. Requested load context was 32,768. Loaded capacity is not actual prompt length or a filled-context test. Actual input token counts are absent. Native prefill throughput/duration are unknown; keep PP and new-input-token fields blank. Method: one stream, max_tokens 600, temperature 0.7, reasoning_effort="none", three runs per prompt. A short Hallo warm-up precedes the measured prompts. The script includes emitted content and reasoning chunks; requested reasoning-off does not establish that reasoning was disabled in every backend/model. The harness computes (tokens - 1) / (stream_end - first_nonempty_content_time). It uses API completion_tokens when available and nonzero; otherwise it falls back to the number of nonempty streamed chunks. Raw records contain tokens, chunks, TTFT and generated text, but not the original usage object, finish_reason or total duration. Treat tokens as recorded output counts, with this usage/chunk fallback caveat. Chunk counts need not equal tokenizer token counts. Rates are the saved rounded client rates, not recalculated from TTFT. TTFT is end-to-end client latency, not native prefill duration. Prompts repeat identically across three repetitions; no nonce or explicit cache clear appears in the harness. Prefix-cache state/cached-token counts are unknown. These are not verified cold-prefill runs. A record of 600 equals the requested cap numerically, but finish_reason is not available. KV dtype, batching, microbatching, thread counts and exact GPU allocation are unreported. All 15 runs follow, in source repetition order. Fields: decode tokens/s; recorded output count; nonempty chunks; client TTFT seconds. German prose, Rhine geology (wissen_de), median 7.4 tokens/s: Run 0: 7.4 tokens/s; 424 output count; 423 chunks; TTFT 0.81 s. Run 1: 7.4 tokens/s; 437 output count; 436 chunks; TTFT 0.77 s. Run 2: 7.5 tokens/s; 404 output count; 403 chunks; TTFT 0.49 s. German prose, bead-based immunoassay (beads_de), median 7.4 tokens/s: Run 0: 7.5 tokens/s; 535 output count; 534 chunks; TTFT 0.86 s. Run 1: 7.4 tokens/s; 558 output count; 557 chunks; TTFT 0.67 s. Run 2: 7.4 tokens/s; 560 output count; 559 chunks; TTFT 0.53 s. English prose, Ruhr economy (prosa_en), median 7.5 tokens/s: Run 0: 7.6 tokens/s; 356 output count; 355 chunks; TTFT 1.22 s. Run 1: 7.4 tokens/s; 371 output count; 370 chunks; TTFT 0.85 s. Run 2: 7.5 tokens/s; 356 output count; 355 chunks; TTFT 0.68 s. Python ISO-8601 duration parser (code_py), median 7.6 tokens/s: Run 0: 7.6 tokens/s; 351 output count; 346 chunks; TTFT 1.04 s. Run 1: 7.6 tokens/s; 353 output count; 348 chunks; TTFT 0.5 s. Run 2: 7.7 tokens/s; 389 output count; 382 chunks; TTFT 0.5 s. JSON antigen/disease extraction (json_extr), median 7.7 tokens/s: Run 0: 7.7 tokens/s; 85 output count; 84 chunks; TTFT 0.98 s. Run 1: 7.7 tokens/s; 85 output count; 84 chunks; TTFT 0.39 s. Run 2: 7.6 tokens/s; 85 output count; 84 chunks; TTFT 0.36 s. Quality scope: The README reports the JSON task passed 3 of 3 automatic checks for this model. Prose/code quality was not generally graded. Throughput is not a quality score.