Rankings

Explore benchmarks

Filter by model, hardware, workload and quantization. Every score links to the comparisons and experiments behind it.

Clear filters

Top Engines

Top Engines, Overall, Speed score.
#EngineSpeed score out of 100
1100 /100 Score
294 /100 Score
384 /100 Score
Not ranked in this view
Not rankedMagnitude Magnitude Too few results
Not rankedOllama Ollama Too few results
Not rankedLiteRT-LM Google AI Edge Not counted
Not rankedMLX LM Apple (ml-explore) Not counted
Not rankedRapid-MLX Rapid-MLX contributors Not counted
details

Speed Decode speed: tokens generated per second.

Score: 100 is the best result in every comparison an engine entered; 80 is 80% of the best, on average. Only complete experiments with a passed output check count. How rankings work

The score combines counted comparisons. Select one comparison to see its measured values.

† uzu: approximate comparison. The experiments use different effective KV cache precision: fp32, f16, uzu default. The experiments here use different 4-bit schemes: 8da4w, Q4_K_M, mirai-m. The experiments use different effective KV cache precision: bf16, fp32, litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 4-bit in another scheme: mirai-m, where most use Q4_K_M.

† ExecuTorch: approximate comparison. The experiments use different effective KV cache precision: fp32, f16, uzu default. The experiments here use different 4-bit schemes: 8da4w, Q4_K_M, mirai-m. The experiments use different effective KV cache precision: bf16, fp32, litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 4-bit in another scheme: mlx-4w-g64, where most use Q4_K_M. The experiments use different effective KV cache precision: fp32, litert-lm default, f16. The experiments here use different 4-bit schemes: 8da4w, litert-int4, Q4_K_M.

† llama.cpp: approximate comparison. The experiments use different effective KV cache precision: fp32, f16, uzu default. The experiments here use different 4-bit schemes: 8da4w, Q4_K_M, mirai-m. The experiments use different effective KV cache precision: bf16, fp32, litert-lm default, f16, magnitude default, model-dtype, uzu default. The experiments use different effective KV cache precision: fp32, litert-lm default, f16. The experiments here use different 4-bit schemes: 8da4w, litert-int4, Q4_K_M.

Magnitude: Results in 2 of 6 comparisons; a rank needs 3.

Ollama: Results in 2 of 6 comparisons; a rank needs 3.

LiteRT-LM: Its output check failed: 10/12 of the checks the model passes at BF16; needs 11 and the mandatory ones (needle-start, needle-end, continuation-128). Left out, failed by every BF16 configuration: none.

MLX LM: Its output check failed: 10/12 of the checks the model passes at BF16; needs 11 and the mandatory ones (needle-start, needle-end, continuation-128). Left out, failed by every BF16 configuration: none.

Rapid-MLX: Its output check failed: 10/12 of the checks the model passes at BF16; needs 11 and the mandatory ones (needle-start, needle-end, continuation-128). Left out, failed by every BF16 configuration: none.

Coverage

Speed score per device, out of 100: each device ranked on its own.
DesktopMobile
M5 Max 64 GiB macOS 27 Apple M5 MaxOnePlus 13 16 GB Android 16 Qualcomm Snapdragon 8 Elite (SM8750)Pixel 8 8 GB Android 16 Google Tensor G3iPhone 12 4 GB iOS 27 Apple A14 Bionic
ExecuTorch87 out of 100, in 2 of 2 comparisons100 out of 100, in 2 of 2 comparisons100 out of 100, in 1 of 1 comparisons94 out of 100, in 1 of 1 comparisons
llama.cpp82 out of 100, in 2 of 2 comparisons91 out of 100, in 2 of 2 comparisons78 out of 100, in 1 of 1 comparisons83 out of 100, in 1 of 1 comparisons
uzu100 out of 100, in 2 of 2 comparisonsNot measured on this device.Not measured on this device.100 out of 100, in 1 of 1 comparisons
Magnitude84 out of 100, in 2 of 2 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.
Ollama84 out of 100, in 2 of 2 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.
LiteRT-LMNot counted Its output check failed: 10/12 of the checks the model passes at BF16; needs 11 and the mandatory ones (needle-start, needle-end, continuation-128). Left out, failed by every BF16 configuration: none.Not counted Its output check failed: 9/12 of the checks the model passes at BF16; needs 11 and the mandatory ones (needle-start, needle-end, continuation-128). Left out, failed by every BF16 configuration: none.Not counted Its output check failed: 9/12 of the checks the model passes at BF16; needs 11 and the mandatory ones (needle-start, needle-end, continuation-128). Left out, failed by every BF16 configuration: none.Not measured on this device.
MLX LMNot counted Its output check failed: 10/12 of the checks the model passes at BF16; needs 11 and the mandatory ones (needle-start, needle-end, continuation-128). Left out, failed by every BF16 configuration: none.Not measured on this device.Not measured on this device.Not measured on this device.
Rapid-MLXNot counted Its output check failed: 10/12 of the checks the model passes at BF16; needs 11 and the mandatory ones (needle-start, needle-end, continuation-128). Left out, failed by every BF16 configuration: none.Not measured on this device.Not measured on this device.Not measured on this device.

Speed

TTFT ms

Generation tok/s

Speed, LFM2.5 1.2B Instruct on iPhone 12 · 4 GB, 4-bit · Archive 256 input / 128 output · iOS 27
RankEngineDecode and TTFT
1uzu, approximate comparison, see details Mirai45.4 tok/s 1,073 ms TTFT
2ExecuTorch, approximate comparison, see details PyTorch (Meta)42.5 tok/s 2,033 ms TTFT
3llama.cpp, approximate comparison, see details ggml-org37.6 tok/s 987 ms TTFT
details

LFM2.5 1.2B Instruct on iPhone 12 · 4 GB · 4-bit · Archive 256 input / 128 output · iOS 27

Medians with the model warmed; model loading is excluded. Each table uses one matched comparison.

uzu: † Approximate comparison. The experiments use different effective KV cache precision: fp32, f16, uzu default. The experiments here use different 4-bit schemes: 8da4w, Q4_K_M, mirai-m.

ExecuTorch: † Approximate comparison. The experiments use different effective KV cache precision: fp32, f16, uzu default. The experiments here use different 4-bit schemes: 8da4w, Q4_K_M, mirai-m.

llama.cpp: † Approximate comparison. The experiments use different effective KV cache precision: fp32, f16, uzu default. The experiments here use different 4-bit schemes: 8da4w, Q4_K_M, mirai-m.

Memory usage by engine

Memory usage by engine, LFM2.5 1.2B Instruct on iPhone 12 · 4 GB, 4-bit · Archive 256 input / 128 output · iOS 27
RankEngineMemory
Not rankedExecuTorch PyTorch (Meta)Not measured
Not rankedllama.cpp ggml-orgNot measured
Not rankeduzu MiraiNot measured
details

LFM2.5 1.2B Instruct on iPhone 12 · 4 GB · 4-bit · Archive 256 input / 128 output · iOS 27

Memory ranks require comparable, calibrated measurements. Unavailable measurements remain unranked.

ExecuTorch: No peak memory measured. † Approximate comparison. The experiments use different effective KV cache precision: fp32, f16, uzu default. The experiments here use different 4-bit schemes: 8da4w, Q4_K_M, mirai-m.

llama.cpp: No peak memory measured. † Approximate comparison. The experiments use different effective KV cache precision: fp32, f16, uzu default. The experiments here use different 4-bit schemes: 8da4w, Q4_K_M, mirai-m.

uzu: No peak memory measured. † Approximate comparison. The experiments use different effective KV cache precision: fp32, f16, uzu default. The experiments here use different 4-bit schemes: 8da4w, Q4_K_M, mirai-m.

Every comparison

Comparisons in this view: each engine's score in each comparison, out of 100, or why it has none.
ComparisonuzuExecuTorchllama.cppMagnitudeOllamaLiteRT-LMMLX LMRapid-MLX
LFM2.5 1.2B Instruct on iPhone 12 · 4 GB 4-bit · Archive 256 input / 128 output · iOS 27 100, approximate94, approximate83, approximateNot runNot runNot runNot runNot run
LFM2.5 1.2B Instruct on M5 Max · 64 GiB 4-bit · Archive 2048 input / 128 output · macOS 27 100, approximate86, approximate83, approximate85, approximate84, approximateNot countedNot countedNot counted
LFM2.5 1.2B Instruct on M5 Max · 64 GiB 4-bit · Archive 256 input / 128 output · macOS 27 100, approximate87, approximate81, approximate82, approximate84, approximateNot countedNot countedNot counted
LFM2.5 1.2B Instruct on OnePlus 13 · 16 GB 4-bit · Archive 2048 input / 128 output · Android 16 Not run100, approximate99, approximateNot runNot runNot countedNot runNot run
LFM2.5 1.2B Instruct on OnePlus 13 · 16 GB 4-bit · Archive 256 input / 128 output · Android 16 Not run100, approximate83, approximateNot runNot runNot countedNot runNot run
LFM2.5 1.2B Instruct on Pixel 8 · 8 GB 4-bit · Archive 256 input / 128 output · Android 16 Not run100, approximate78, approximateNot runNot runNot countedNot runNot run
LFM2.5 1.2B Instruct on iPhone 12 · 4 GB 4-bit · Archive 2048 input / 128 output · iOS 27 Not counted: fewer than two scored enginesNot countedNot countedMeasured, approximateNot runNot runNot runNot runNot run
LFM2.5 1.2B Instruct on M5 Max · 64 GiB 4-bit · 2048 input / 128 output · macOS 27 Not counted: fewer than two scored enginesNot runNot runNot countedNot runNot runNot runNot countedNot run
LFM2.5 1.2B Instruct on M5 Max · 64 GiB 4-bit · 256 input / 128 output · macOS 27 Not counted: fewer than two scored enginesNot runNot runMeasured, approximateNot runNot runNot runNot countedNot run
LFM2.5 1.2B Instruct on Pixel 8 · 8 GB 4-bit · Archive 2048 input / 128 output · Android 16 Not counted: fewer than two scored enginesNot runNot countedNot countedNot runNot runNot countedNot runNot run

Each engine’s Speed score in each matched comparison, out of 100, or why it has none. How rankings work