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
1MLX LM, an approximate comparison, see details Apple (ml-explore) 100 /100 Score
2Rapid-MLX, an approximate comparison, see details Rapid-MLX contributors 100 /100 Score
3Ollama, an approximate comparison, see details Ollama 96 /100 Score
4llama.cpp, an approximate comparison, see details ggml-org 89 /100 Score
Not ranked in this view
Not rankedMagnitude Magnitude Unsupported
Not rankeduzu Mirai Unsupported
Not rankedLiteRT-LM Google AI Edge 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

† MLX LM: approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

† Rapid-MLX: approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

† Ollama: approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

† llama.cpp: approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

Magnitude: Unsupported in 1 comparison. Magnitude's chat API renders LFM2.5 2.6B's own template, which always opens a thinking block, so it streams reasoning and can't run the closed-think prompts (measured in qualification).

uzu: Unsupported in 1 comparison. uzu can't run the frozen prompts: Error: uzu counted 23 input tokens, not the frozen prompt's 26 (measured in qualification).

LiteRT-LM: Its output check failed: 7/9 of the checks the model passes at BF16; needs 8 and the mandatory ones (needle-start, continuation-128). Left out, failed by every BF16 configuration: color, sort, needle-end.

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.cpp89 out of 100, in 1 of 1 comparisonsNo rival yet No other engine has a result in the same comparison yet.Failed Failed in 1 comparison. No trial was valid.Unsupported Unsupported in 1 comparison. llama.cpp errored on the arithmetic check in qualification: The app failed: llama_decode(token 1) returned -3
MLX LM100 out of 100, in 1 of 1 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.
Ollama96 out of 100, in 1 of 1 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.
Rapid-MLX100 out of 100, in 1 of 1 comparisonsNot measured on this device.Not measured on this device.Not measured on this device.
LiteRT-LMNot counted Its output check failed: 7/9 of the checks the model passes at BF16; needs 8 and the mandatory ones (needle-start, continuation-128). Left out, failed by every BF16 configuration: color, sort, needle-end.Not counted Partial: 2 of 6 planned trials were valid.Failed Failed in 1 comparison. No trial was valid.Not measured on this device.
uzuUnsupported Unsupported in 1 comparison. uzu can't run the frozen prompts: Error: uzu counted 23 input tokens, not the frozen prompt's 26 (measured in qualification).Not measured on this device.Not measured on this device.Unsupported Unsupported in 1 comparison. uzu can't run the frozen prompts: Error: uzu counted 23 input tokens, not the frozen prompt's 26 (measured in qualification).
MagnitudeUnsupported Unsupported in 1 comparison. Magnitude's chat API renders LFM2.5 2.6B's own template, which always opens a thinking block, so it streams reasoning and can't run the closed-think prompts (measured in qualification).Not measured on this device.Not measured on this device.Not measured on this device.

Speed

TTFT ms

Generation tok/s

Speed, LFM2.5 2.6B on M5 Max · 64 GiB, 8-bit · Archive 2048 input / 128 output · macOS 27
RankEngineDecode and TTFT
1MLX LM, approximate comparison, see details Apple (ml-explore)160.4 tok/s 221 ms TTFT
2Rapid-MLX, approximate comparison, see details Rapid-MLX contributors160.0 tok/s 242 ms TTFT
3Ollama, approximate comparison, see details Ollama153.6 tok/s 283 ms TTFT
4llama.cpp, approximate comparison, see details ggml-org142.4 tok/s 298 ms TTFT
Not rankedLiteRT-LM Google AI EdgeNot counted
Not rankedMagnitude MagnitudeUnsupported
Not rankeduzu MiraiUnsupported
details

LFM2.5 2.6B on M5 Max · 64 GiB · 8-bit · Archive 2048 input / 128 output · macOS 27

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

MLX LM: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

Rapid-MLX: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

Ollama: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

llama.cpp: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

LiteRT-LM: Its output check failed: 7/9 of the checks the model passes at BF16; needs 8 and the mandatory ones (needle-start, continuation-128). Left out, failed by every BF16 configuration: color, sort, needle-end. † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

Magnitude: Magnitude's chat API renders LFM2.5 2.6B's own template, which always opens a thinking block, so it streams reasoning and can't run the closed-think prompts (measured in qualification).

uzu: uzu can't run the frozen prompts: Error: uzu counted 23 input tokens, not the frozen prompt's 26 (measured in qualification).

Memory usage by engine

Memory usage by engine, LFM2.5 2.6B on M5 Max · 64 GiB, 8-bit · Archive 2048 input / 128 output · macOS 27
RankEngineMemory
1Ollama, approximate comparison, see details Ollama2.97 GiB
2llama.cpp, approximate comparison, see details ggml-org2.98 GiB
3MLX LM, approximate comparison, see details Apple (ml-explore)4.01 GiB
4Rapid-MLX, approximate comparison, see details Rapid-MLX contributors4.14 GiB
Not rankedLiteRT-LM Google AI EdgeNot counted
Not rankedMagnitude MagnitudeUnsupported
Not rankeduzu MiraiUnsupported
details

LFM2.5 2.6B on M5 Max · 64 GiB · 8-bit · Archive 2048 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, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

llama.cpp: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

MLX LM: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

Rapid-MLX: † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

LiteRT-LM: Its output check failed: 7/9 of the checks the model passes at BF16; needs 8 and the mandatory ones (needle-start, continuation-128). Left out, failed by every BF16 configuration: color, sort, needle-end. † Approximate comparison. The experiments use different effective KV cache precision: litert-lm default, f16, model-dtype. The experiments here use different 8-bit schemes: litert-int8, Q8_0, mlx-affine-8bit-g64.

Magnitude: Magnitude's chat API renders LFM2.5 2.6B's own template, which always opens a thinking block, so it streams reasoning and can't run the closed-think prompts (measured in qualification).

uzu: uzu can't run the frozen prompts: Error: uzu counted 23 input tokens, not the frozen prompt's 26 (measured in qualification).

Every comparison

Comparisons in this view: each engine's score in each comparison, out of 100, or why it has none.
ComparisonMLX LMRapid-MLXOllamallama.cppMagnitudeuzuLiteRT-LM
LFM2.5 2.6B on M5 Max · 64 GiB 8-bit · Archive 2048 input / 128 output · macOS 27 100, approximate100, approximate96, approximate89, approximateUnsupportedUnsupportedNot counted

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