Rankings

Explore benchmarks

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

Top Engines

Top Engines, Desktop, Speed score.
#EngineSpeed score out of 100
199 /100 Score
295 /100 Score
395 /100 Score
495 /100 Score
577 /100 Score
676 /100 Score
725 /100 Score
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: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: mirai-l, where most use Q8_0.

† MLX LM: approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: mlx-affine-8bit-g64, where most use Q8_0.

† Magnitude: approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its phases were timed by the harness clock around the engine API, so they include transport that engine counters leave out.

† Rapid-MLX: approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: mlx-affine-8bit-g64, where most use Q8_0.

† Ollama: approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default.

† llama.cpp: approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default.

† LiteRT-LM: approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: litert-int8, where most use Q8_0.

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
llama.cpp76 out of 100, in 2 of 2 comparisons100 out of 100, in 2 of 2 comparisons90 out of 100, in 2 of 2 comparisons68 out of 100, in 2 of 2 comparisons
LiteRT-LM25 out of 100, in 2 of 2 comparisons70 out of 100, in 2 of 2 comparisons96 out of 100, in 2 of 2 comparisonsNot measured on this device.
uzu99 out of 100, in 2 of 2 comparisonsNot measured on this device.Not measured on this device.100 out of 100, in 2 of 2 comparisons
Magnitude95 out of 100, in 2 of 2 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.
MLX LM95 out of 100, in 2 of 2 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.
Ollama77 out of 100, in 2 of 2 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.
Rapid-MLX95 out of 100, in 2 of 2 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.

Speed

details

LFM2.5 350M on M5 Max · 64 GiB · 8-bit · Archive 256 input / 128 output · macOS 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: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: mirai-l, where most use Q8_0.

Magnitude: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its phases were timed by the harness clock around the engine API, so they include transport that engine counters leave out.

Ollama: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default.

llama.cpp: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default.

MLX LM: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: mlx-affine-8bit-g64, where most use Q8_0.

Rapid-MLX: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: mlx-affine-8bit-g64, where most use Q8_0.

LiteRT-LM: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: litert-int8, where most use Q8_0.

Memory usage by engine

details

LFM2.5 350M on M5 Max · 64 GiB · 8-bit · Archive 256 input / 128 output · macOS 27

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

Ollama: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default.

llama.cpp: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default.

uzu: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: mirai-l, where most use Q8_0.

MLX LM: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: mlx-affine-8bit-g64, where most use Q8_0.

LiteRT-LM: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: litert-int8, where most use Q8_0.

Rapid-MLX: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its weights are 8-bit in another scheme: mlx-affine-8bit-g64, where most use Q8_0.

Magnitude: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, magnitude default, model-dtype, uzu default. Its phases were timed by the harness clock around the engine API, so they include transport that engine counters leave out.

Every comparison

Comparisons in this view: each engine's score in each comparison, out of 100, or why it has none.
ComparisonuzuMLX LMMagnitudeRapid-MLXOllamallama.cppLiteRT-LM
LFM2.5 350M on M5 Max · 64 GiB 8-bit · Archive 2048 input / 128 output · macOS 27 100, approximate90, approximate96, approximate92, approximate77, approximate76, approximate26, approximate
LFM2.5 350M on M5 Max · 64 GiB 8-bit · Archive 256 input / 128 output · macOS 27 97, approximate100, approximate93, approximate98, approximate77, approximate75, approximate24, approximate

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