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
1100 /100 Score
295 /100 Score
382 /100 Score
476 /100 Score
572 /100 Score
Not ranked in this view
Not rankedExecuTorch PyTorch (Meta) Too few results
Not rankedRapid-MLX Rapid-MLX contributors Too few results
Not rankedLiteRT-LM Google AI Edge No rival yet
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.

† MLX LM: approximate comparison. 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-affine-4bit-g64, where most use Q4_K_M. The experiments use different effective KV cache precision: f16, model-dtype.

† uzu: approximate comparison. 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.

† Magnitude: approximate comparison. The experiments use different effective KV cache precision: bf16, fp32, 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: bf16, fp32, litert-lm default, f16, magnitude default, model-dtype, uzu default. The experiments use different effective KV cache precision: f16, model-dtype.

† llama.cpp: approximate comparison. 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: f16, model-dtype.

ExecuTorch: Results in 1 of 3 comparisons; a rank needs 2.

Rapid-MLX: Results in 1 of 3 comparisons; a rank needs 2.

LiteRT-LM: No other engine has a result in the same comparison yet.

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.cpp72 out of 100, in 3 of 3 comparisons83 out of 100, in 1 of 1 comparisons78 out of 100, in 1 of 1 comparisons60 out of 100, in 2 of 2 comparisons
ExecuTorchNot ranked. Results in 1 of 3 comparisons; a rank needs 2.100 out of 100, in 1 of 1 comparisons100 out of 100, in 1 of 1 comparisonsNot ranked. Results in 1 of 2 comparisons; a rank needs 2.
uzu95 out of 100, in 2 of 3 comparisonsNot measured on this device.Not measured on this device.100 out of 100, in 2 of 2 comparisons
Magnitude82 out of 100, in 2 of 3 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.
MLX LM100 out of 100, in 2 of 3 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.
Ollama76 out of 100, in 3 of 3 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.
LiteRT-LMNo rival yet No other engine has a result in the same comparison yet.No rival yet No other engine has a result in the same comparison yet.No rival yet No other engine has a result in the same comparison yet.Not measured on this device.
Rapid-MLXNot ranked. Results in 1 of 3 comparisons; a rank needs 2.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 M5 Max · 64 GiB, 4-bit · Archive 256 input / 128 output · macOS 27
RankEngineDecode and TTFT
1uzu, approximate comparison, see details Mirai540.7 tok/s 75.2 ms TTFT
2ExecuTorch, approximate comparison, see details PyTorch (Meta)472.1 tok/s 52.3 ms TTFT
3Ollama, approximate comparison, see details Ollama454.6 tok/s 30.6 ms TTFT
4Magnitude, approximate comparison, see details Magnitude444.5 tok/s 26.6 ms TTFT
5llama.cpp, approximate comparison, see details ggml-org438.1 tok/s 34.2 ms TTFT
Not rankedLiteRT-LM Google AI EdgeNot counted
Not rankedMLX LM Apple (ml-explore)Not counted
Not rankedRapid-MLX Rapid-MLX contributorsNot counted
details

LFM2.5 1.2B Instruct on M5 Max · 64 GiB · 4-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: 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: 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.

Ollama: † Approximate comparison. The experiments use different effective KV cache precision: bf16, fp32, litert-lm default, f16, magnitude default, model-dtype, uzu default.

Magnitude: † Approximate comparison. The experiments use different effective KV cache precision: bf16, fp32, 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.

llama.cpp: † Approximate comparison. The experiments use different effective KV cache precision: bf16, fp32, litert-lm default, f16, magnitude default, model-dtype, uzu default.

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. † Approximate comparison. 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: litert-int4, where most use Q4_K_M.

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. † Approximate comparison. 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-affine-4bit-g64, where most use Q4_K_M.

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. † Approximate comparison. 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-affine-4bit-g64, where most use Q4_K_M.

Memory usage by engine

Memory usage by engine, LFM2.5 1.2B Instruct on M5 Max · 64 GiB, 4-bit · Archive 256 input / 128 output · macOS 27
RankEngineMemory
1uzu, approximate comparison, see details Mirai0.90 GiB
2Ollama, approximate comparison, see details Ollama0.93 GiB
3llama.cpp, approximate comparison, see details ggml-org0.93 GiB
4Magnitude, approximate comparison, see details Magnitude1.56 GiB
Not rankedExecuTorch PyTorch (Meta)Not measured
Not rankedLiteRT-LM Google AI EdgeNot counted
Not rankedMLX LM Apple (ml-explore)Not counted
Not rankedRapid-MLX Rapid-MLX contributorsNot counted
details

LFM2.5 1.2B Instruct on M5 Max · 64 GiB · 4-bit · Archive 256 input / 128 output · macOS 27

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

uzu: † Approximate comparison. 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.

Ollama: † Approximate comparison. The experiments use different effective KV cache precision: bf16, fp32, litert-lm default, f16, magnitude default, model-dtype, uzu default.

llama.cpp: † Approximate comparison. The experiments use different effective KV cache precision: bf16, fp32, litert-lm default, f16, magnitude default, model-dtype, uzu default.

Magnitude: † Approximate comparison. The experiments use different effective KV cache precision: bf16, fp32, 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.

ExecuTorch: No peak memory measured. † Approximate comparison. 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.

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. † Approximate comparison. 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: litert-int4, where most use Q4_K_M.

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. † Approximate comparison. 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-affine-4bit-g64, where most use Q4_K_M.

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. † Approximate comparison. 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-affine-4bit-g64, where most use Q4_K_M.

Every comparison

Comparisons in this view: each engine's score in each comparison, out of 100, or why it has none.
ComparisonMLX LMuzuMagnitudeOllamallama.cppExecuTorchRapid-MLXLiteRT-LM
LFM2.5 1.2B Instruct on M5 Max · 64 GiB 4-bit · Archive 256 input / 128 output · macOS 27 Not counted100, approximate82, approximate84, approximate81, approximate87, approximateNot countedNot counted
LFM2.5 350M on M5 Max · 64 GiB 4-bit · Archive 256 input / 128 output · macOS 27 100, approximate90, approximate82, approximate67, approximate62, approximateNot counted100, approximateNot counted
Qwen3.8 27B on M5 Max · 64 GiB 4-bit · Archive 256 input / 128 output · macOS 27 100, approximateUnsupportedUnsupported78, approximate75, approximateNot runNot runNot run
LFM2.5 2.6B on M5 Max · 64 GiB 4-bit · Archive 256 input / 128 output · macOS 27 Not counted: fewer than two scored enginesNot countedUnsupportedUnsupportedNot countedNot countedNot runNot countedMeasured, approximate

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