Models

Loading community benchmarks…

← All models

Community benchmarks

Llama 3.2 3B Instruct

Submitted benchmarks for Llama 3.2 3B Instruct. Browse individual runs and their measurement settings, or open the report JSON for full benchmark details.

Compare chips Compare runtimes Add your benchmark

Submitted runs: 26
Chips tested: 8
Contributors: 2

Model identity

Grouped as meta-llama/Llama-3.2-3B-Instruct. The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution. · Nothing in this report ties it to a published model file. Reports from CLI 0.1.0 record only a model name and a file hash, so the run is grouped by the name it reported until it is reconciled.. Artifact verification confirms published model bytes, not that a community-submitted benchmark was honestly executed. Hover or tap a label for what it means.

This model group collects runs reported as basecompute/Llama-3.2-3B-Instruct, Llama 3.2 3B Instruct, Llama-3.2-3B-Instruct, Llama-3.2-3B-Instruct-cuda-q4mix, Llama-3.2-3B-Instruct-cuda-q8, Llama-3.2-3B-Instruct-Q4, Llama-3.2-3B-Instruct-Q8. Individual results keep the package name and format they were submitted with.

View on Hugging Face ↗

Published files benchmarked

basecompute/Llama-3.2-3B-Instruct/Llama-3.2-3B-Instruct-Q4.base (opens in a new tab)BaseRTQ4· 8 runs
basecompute/Llama-3.2-3B-Instruct/Llama-3.2-3B-Instruct-Q8.base (opens in a new tab)BaseRTQ4· 7 runs
unsloth/Llama-3.2-3B-Instruct-GGUF/Llama-3.2-3B-Instruct-Q4_K_M.gguf (opens in a new tab)llama.cppQ4_K_M· 3 runs

Benchmark results

Peak throughput is the highest reported value and may come from different runs. Compare model, quantisation, backend, and token counts before drawing conclusions. Model verification identifies published artifact bytes; community benchmark execution is not remotely attested.

Individual runs for Llama 3.2 3B Instruct. Throughput is in tokens per second; dates use your browser timezone. Hover a date for the exact local time.
Model / runtime / format	Chip / backend	Decode	Prefill	Peak process memory	Contributor

Llama-3.2-3B-Instruct-Q4 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M4 Pro Metal	95.0 tok/s TG128 @ 1 ctx	943 tok/s PP512	3,089.2 MiB	basecompute	20 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct-Q8 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M4 Pro Metal	71.2 tok/s TG128 @ 1 ctx	943 tok/s PP512	4,772.4 MiB	basecompute	20 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct-Q4 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M3 Ultra Metal	195.4 tok/s TG128 @ 1 ctx	3,241 tok/s PP512	3,085.0 MiB	basecompute	20 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct-Q8 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M3 Ultra Metal	154.4 tok/s TG128 @ 1 ctx	3,316 tok/s PP512	4,768.0 MiB	basecompute	20 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct-Q4 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M1 Max Metal	110.7 tok/s TG128 @ 1 ctx	1,229 tok/s PP512	3,091.7 MiB	basecompute	20 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct-Q8 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M1 Max Metal	88.4 tok/s TG128 @ 1 ctx	1,240 tok/s PP512	4,774.1 MiB	basecompute	20 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct-Q8 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M4 Max Metal	133.1 tok/s TG128 @ 1 ctx	2,218 tok/s PP512	4,770.8 MiB	basecompute	19 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct-Q4 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M4 Max Metal	178.8 tok/s TG128 @ 1 ctx	2,202 tok/s PP512	3,087.5 MiB	basecompute	19 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct-cuda-q8 BaseRTQ4 Nothing in this report ties it to a published model file. Reports from CLI 0.1.0 record only a model name and a file hash, so the run is grouped by the name it reported until it is reconciled.	NVIDIA GB10 CUDA	69.4 tok/s TG128 @ 1 ctx	14,988 tok/s PP512	3,611.8 MiB	basecompute	19 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct-Q8 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	NVIDIA GB10 CUDA	68.7 tok/s TG128 @ 1 ctx	14,871 tok/s PP512	3,611.6 MiB	basecompute	19 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct-Q4 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	NVIDIA GB10 CUDA	89.7 tok/s TG128 @ 1 ctx	14,073 tok/s PP512	2,891.3 MiB	basecompute	19 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct-cuda-q4mix BaseRTQ4 Nothing in this report ties it to a published model file. Reports from CLI 0.1.0 record only a model name and a file hash, so the run is grouped by the name it reported until it is reconciled.	NVIDIA GB10 CUDA	110.5 tok/s TG128 @ 1 ctx	14,651 tok/s PP512	2,444.5 MiB	basecompute	19 Sep 2026	View Benchmark
Llama 3.2 3B Instruct llama.cppQ8_0 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	AMD Radeon RX 7900 XT (RADV NAVI31) Vulkan	167.5 tok/s TG128	5,846 tok/s PP512	3,394.5 MiB	arki05	11 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct llama.cppQ4_K_M The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	AMD Radeon RX 7900 XT (RADV NAVI31) Vulkan	223.4 tok/s TG128	5,721 tok/s PP512	437.0 MiB	arki05	11 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct llama.cppQ4_K_M The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	AMD Radeon RX 7900 XT (RADV NAVI31) Vulkan	229.7 tok/s TG128	5,737 tok/s PP512	442.4 MiB	arki05	11 Sep 2026	View Benchmark
Llama 3.2 3B Instruct llama.cppQ8_0 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	AMD Radeon RX 7900 XT ROCm	142.3 tok/s TG128	7,210 tok/s PP512	5,102.3 MiB	arki05	11 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct llama.cppQ4_K_M The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	AMD Radeon RX 7900 XT ROCm	171.9 tok/s TG128	6,832 tok/s PP512	3,804.4 MiB	arki05	11 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct llama.cppQ4_K_M The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	AMD Radeon RX 7900 XT ROCm	172.0 tok/s TG128	6,980 tok/s PP512	3,764.9 MiB	arki05	11 Sep 2026	View Benchmark
basecompute/Llama-3.2-3B-Instruct BaseRTQ4· default-q8 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M5 Pro Metal	79.5 tok/s TG128	4,343 tok/s PP512	4,779.3 MiB	arki05	11 Sep 2026	View Benchmark
basecompute/Llama-3.2-3B-Instruct BaseRTQ4· default-q4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M5 Pro Metal	106.8 tok/s TG128	4,297 tok/s PP512	3,096.1 MiB	arki05	11 Sep 2026	View Benchmark
Llama 3.2 3B Instruct llama.cppQ8_0 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M5 Pro BLAS + Metal	76.6 tok/s TG128	3,516 tok/s PP512	5,237.3 MiB	arki05	11 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct llama.cppQ4_K_M The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M5 Pro BLAS + Metal	115.2 tok/s TG128	3,315 tok/s PP512	3,940.1 MiB	arki05	11 Sep 2026	View Benchmark
Llama-3.2-3B-Instruct llama.cppQ4_K_M The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M5 Pro BLAS + Metal	117.7 tok/s TG128	3,318 tok/s PP512	3,904.2 MiB	arki05	11 Sep 2026	View Benchmark
basecompute/Llama-3.2-3B-Instruct BaseRTQ4· default-q4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M5 Pro Metal	94.6 tok/s TG128	3,845 tok/s PP512	3,096.0 MiB	arki05	11 Sep 2026	View Benchmark
basecompute/Llama-3.2-3B-Instruct BaseRTQ4· default-q8 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M5 Pro Metal	79.5 tok/s TG128	4,344 tok/s PP512	4,779.6 MiB	arki05	11 Sep 2026	View Benchmark