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 | 100 /100 Score |
| ||
| 2 | llama.cpp | 97 /100 Score |
| ||
| Not ranked in this view | ||
| Not ranked | Magnitude | Unsupported |
| Not ranked | ZML | 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
The score combines counted comparisons. Select one comparison to see its measured values.
Magnitude: Unsupported in 2 comparisons. 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).
ZML: Unsupported in 2 comparisons. zml can't run the frozen prompts: Error: LLMD streamed reasoning before its answer; the harness can't time it apart from the answer (measured in qualification).
Coverage
| Desktop | |
|---|---|
| M5 Max 64 GiB macOS 27 Apple M5 Max | |
| llama.cpp | 972/2 out of 100, in 2 of 2 comparisons |
| Ollama | 1002/2 out of 100, in 2 of 2 comparisons |
| Magnitude | Unsupported Unsupported in 2 comparisons. 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). |
| ZML | Unsupported Unsupported in 2 comparisons. zml can't run the frozen prompts: Error: LLMD streamed reasoning before its answer; the harness can't time it apart from the answer (measured in qualification). |
Speed
| Rank | Engine | Decode and TTFT |
|---|---|---|
| 1 | Ollama | 91.7 tok/s 57.0 ms TTFT |
| 2 | llama.cpp | 88.1 tok/s 68.8 ms TTFT |
| Not ranked | Magnitude | Unsupported |
| Not ranked | ZML | Unsupported |
details
LFM2.5 2.6B on M5 Max · 64 GiB · bf16 · Archive 256 input / 128 output · macOS 27
Medians with the model warmed; model loading is excluded. Each table uses one matched comparison.
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).
ZML: zml can't run the frozen prompts: Error: LLMD streamed reasoning before its answer; the harness can't time it apart from the answer (measured in qualification).
Memory usage by engine
| Rank | Engine | Memory |
|---|---|---|
| 1 | Ollama | 5.31 GiB |
| 2 | llama.cpp | 5.31 GiB |
| Not ranked | Magnitude | Unsupported |
| Not ranked | ZML | Unsupported |
details
LFM2.5 2.6B on M5 Max · 64 GiB · bf16 · Archive 256 input / 128 output · macOS 27
Memory ranks require comparable, calibrated measurements. Unavailable measurements remain unranked.
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).
ZML: zml can't run the frozen prompts: Error: LLMD streamed reasoning before its answer; the harness can't time it apart from the answer (measured in qualification).
Every comparison
| Comparison | Ollama | llama.cpp | Magnitude | ZML |
|---|---|---|---|---|
| LFM2.5 2.6B on M5 Max · 64 GiB bf16 · Archive 2048 input / 128 output · macOS 27 | 100 | 97 | Unsupported | Unsupported |
| LFM2.5 2.6B on M5 Max · 64 GiB bf16 · Archive 256 input / 128 output · macOS 27 | 100 | 96 | Unsupported | Unsupported |
Each engine’s Speed score in each matched comparison, out of 100, or why it has none. How rankings work