Explore benchmarks
Filter by model, hardware, workload and quantization. Every score links to the comparisons and experiments behind it.
Top Engines
| # | Engine | Speed score out of 100 |
|---|---|---|
| 1 | Ollama†, an approximate comparison, see details | 100 /100 Score |
| 2 | llama.cpp†, an approximate comparison, see details | 98 /100 Score |
| 3 | MLX LM†, an approximate comparison, see details | 97 /100 Score |
| Not ranked in this view | ||
| Not ranked | Magnitude | Unsupported |
| Not ranked | uzu | Unsupported |
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
† Ollama: approximate comparison. The experiments use different effective KV cache precision: f16, model-dtype.
† llama.cpp: approximate comparison. The experiments use different effective KV cache precision: f16, model-dtype.
† MLX LM: approximate comparison. The experiments use different effective KV cache precision: f16, model-dtype. Its weights are 8-bit in another scheme: mlx-affine-8bit-g64, where most use Q8_0.
Magnitude: Unsupported in 1 comparison. Magnitude 0.2.17 reports this GGUF as Unsupported (magnitude catalog inspect)
uzu: Unsupported in 1 comparison. uzu 0.6.17 renders its own Qwen3.8 chat template: 67 input tokens for the 27-token canonical check prompt, so it can't run the frozen prompts (measured in qualification).
Coverage
| Desktop | |
|---|---|
| M5 Max 64 GiB macOS 27 Apple M5 Max | |
| llama.cpp | 981/1 out of 100, in 1 of 1 comparisons |
| MLX LM | 971/1 out of 100, in 1 of 1 comparisons |
| Ollama | 1001/1 out of 100, in 1 of 1 comparisons |
| Magnitude | Unsupported Unsupported in 1 comparison. Magnitude 0.2.17 reports this GGUF as Unsupported (magnitude catalog inspect) |
| uzu | Unsupported Unsupported in 1 comparison. uzu 0.6.17 renders its own Qwen3.8 chat template: 67 input tokens for the 27-token canonical check prompt, so it can't run the frozen prompts (measured in qualification). |
Speed
| Rank | Engine | Decode and TTFT |
|---|---|---|
| 1 | Ollama†, approximate comparison, see details | 18.9 tok/s 2,748 ms TTFT |
| 2 | llama.cpp†, approximate comparison, see details | 18.5 tok/s 2,996 ms TTFT |
| 3 | MLX LM†, approximate comparison, see details | 18.3 tok/s 2,403 ms TTFT |
| Not ranked | Magnitude | Unsupported |
| Not ranked | uzu | Unsupported |
details
Qwen3.8 27B 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.
Ollama: † Approximate comparison. The experiments use different effective KV cache precision: f16, model-dtype.
llama.cpp: † Approximate comparison. The experiments use different effective KV cache precision: f16, model-dtype.
MLX LM: † Approximate comparison. The experiments use different effective KV cache precision: f16, model-dtype. Its weights are 8-bit in another scheme: mlx-affine-8bit-g64, where most use Q8_0.
Magnitude: Magnitude 0.2.17 reports this GGUF as Unsupported (magnitude catalog inspect)
uzu: uzu 0.6.17 renders its own Qwen3.8 chat template: 67 input tokens for the 27-token canonical check prompt, so it can't run the frozen prompts (measured in qualification).
Memory usage by engine
| Rank | Engine | Memory |
|---|---|---|
| 1 | Ollama†, approximate comparison, see details | 27.63 GiB |
| 2 | llama.cpp†, approximate comparison, see details | 27.65 GiB |
| 3 | MLX LM†, approximate comparison, see details | 30.39 GiB |
| Not ranked | Magnitude | Unsupported |
| Not ranked | uzu | Unsupported |
details
Qwen3.8 27B 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: f16, model-dtype.
llama.cpp: † Approximate comparison. The experiments use different effective KV cache precision: f16, model-dtype.
MLX LM: † Approximate comparison. The experiments use different effective KV cache precision: f16, model-dtype. Its weights are 8-bit in another scheme: mlx-affine-8bit-g64, where most use Q8_0.
Magnitude: Magnitude 0.2.17 reports this GGUF as Unsupported (magnitude catalog inspect)
uzu: uzu 0.6.17 renders its own Qwen3.8 chat template: 67 input tokens for the 27-token canonical check prompt, so it can't run the frozen prompts (measured in qualification).
Every comparison
| Comparison | Ollama | llama.cpp | MLX LM | Magnitude | uzu |
|---|---|---|---|---|---|
| Qwen3.8 27B on M5 Max · 64 GiB 8-bit · Archive 2048 input / 128 output · macOS 27 | 100†, approximate | 98†, approximate | 97†, approximate | Unsupported | Unsupported |
Each engine’s Speed score in each matched comparison, out of 100, or why it has none. How rankings work