Compare

Loading community benchmarks…

Community benchmarks

Compare

Put two chips, runtimes, or runtime versions side by side. Everything else is held fixed or called out, so a ratio is worth exactly what the matched runs say.

Add your benchmark

Runs compared: 7
Like-for-like pair: 1
Chips available: 6

Compare

Side A

Side B

Model

Runtime

Runtime version

Quantisation

Backend

Prefill size

Decode size

Cooldown

4 runs vs 3 runs · 1 like-for-like pair

Held fixedModel: Llama 3.2 1B InstructPrefill: PP512Decode: TG128

The sides differ in decode workload. Ratios may reflect benchmark protocol rather than the chip.

Chip comparison

NVIDIA GB10 vs Apple M5 Pro

Ratios read side A relative to side B. Geometric mean over 1 like-for-like pair.

Decode: 0.89×-10.6%
Prefill: 3.01×+201%

Side A

NVIDIA GB10

4 runs

1 contributor

Latest 19 Sep 2026

Decode · TG128: 184.5tok/s median
best 261.3 · worst 172.8
Prefill · PP512: 35,119tok/s median
best 36,647 · worst 33,397

Models: Llama-3.2-1B-Instruct-cuda-q4mix, Llama-3.2-1B-Instruct-cuda-q8, Llama-3.2-1B-Instruct-Q4, Llama-3.2-1B-Instruct-Q8
Runtimes: BaseRT
Versions: BaseRT 0.2.6
Quantisations: Q4
Backends: CUDA
Cooldown: off
Conditioning: warmup_only
Decode workload: TG128 @ 1 ctx
Harness schema: basert-benchmark-harness/1

Side B

Apple M5 Pro

3 runs

1 contributor

Latest 11 Sep 2026

Decode · TG128: 206.3tok/s median
best 229.1 · worst 205.3
Prefill · PP512: 11,673tok/s median
best 12,067 · worst 10,528

Models: Llama-3.2-1B-Instruct
Runtimes: BaseRT
Versions: BaseRT 0.2.4
Quantisations: Q4
Backends: Metal
Cooldown: off
Conditioning: warmup_only
Decode workload: TG128
Harness schema: basert-benchmark-harness/1

What differs besides the chip

Facet	NVIDIA GB10	Apple M5 Pro
Model	Llama-3.2-1B-Instruct-cuda-q4mix, Llama-3.2-1B-Instruct-cuda-q8, Llama-3.2-1B-Instruct-Q4, Llama-3.2-1B-Instruct-Q8	Llama-3.2-1B-Instruct
Runtime version	BaseRT 0.2.6	BaseRT 0.2.4
Backend	CUDA	Metal
Decode workload	TG128 @ 1 ctx	TG128

Like-for-like pairs

Each row is a configuration present on both sides. Values are per-cell medians; ratios read NVIDIA GB10 relative to Apple M5 Pro.

Configuration	Decode A	Decode B	Ratio	Prefill A	Prefill B	Ratio	Runs A / B
Llama-3.2-1B-Instruct-cuda-q8 BaseRTQ4	184.5	206.3	0.89×	35,119	11,673	3.01×	4 / 3

Side A · NVIDIA GB10

Top 4 of 4 runs by decode throughput.

Model / format	Chip / backend	Decode	Prefill	Contributor	Date	Report
Llama-3.2-1B-Instruct-cuda-q4mix BaseRT 0.2.6Q4	NVIDIA GB10 CUDA	261.3 TG128 @ 1 ctx	36,647 PP512	basecompute	19 Sep 2026	View Benchmark
Llama-3.2-1B-Instruct-Q4 BaseRT 0.2.6Q4	NVIDIA GB10 CUDA	195.3 TG128 @ 1 ctx	33,397 PP512	basecompute	19 Sep 2026	View Benchmark
Llama-3.2-1B-Instruct-cuda-q8 BaseRT 0.2.6Q4	NVIDIA GB10 CUDA	173.7 TG128 @ 1 ctx	34,296 PP512	basecompute	19 Sep 2026	View Benchmark
Llama-3.2-1B-Instruct-Q8 BaseRT 0.2.6Q4	NVIDIA GB10 CUDA	172.8 TG128 @ 1 ctx	35,941 PP512	basecompute	19 Sep 2026	View Benchmark

Side B · Apple M5 Pro

Top 3 of 3 runs by decode throughput.

Model / format	Chip / backend	Decode	Prefill	Contributor	Date	Report
basecompute/Llama-3.2-1B-Instruct BaseRT 0.2.4Q4	Apple M5 Pro Metal	229.1 TG128	11,673 PP512	arki05	11 Sep 2026	View Benchmark
basecompute/Llama-3.2-1B-Instruct BaseRT 0.2.4Q4	Apple M5 Pro Metal	206.3 TG128	10,528 PP512	arki05	11 Sep 2026	View Benchmark
basecompute/Llama-3.2-1B-Instruct BaseRT 0.2.4Q4	Apple M5 Pro Metal	205.3 TG128	12,067 PP512	arki05	11 Sep 2026	View Benchmark