63.6
% saved ↑
105
ttft ms
25.8
Peak memory GB
120
Trials
Validity: valid
The checks LemonMatrix ran while measuring; an invalid run is shown but never ranked.
- ✓ Warm-up discarded
- ✓ Thermal OK
- ✓ Exclusive run
- ✓ No model reload
- TTFT, prefill and decode cover the 78 of 120 requests served locally
- Lemonade does not time cloud requests
- Lemonade doesn't report power draw for this instance; energy metrics omitted
- exclusive_run verified via /api/v1/health polling (no other loaded model was busy)
Trials
Each measured trial after warm-up, from the run's trial sidecar. The summary above is computed from these.
| # | ttft ms | decode tokens per sec | prompt tokens | output tokens | cache tokens | vram gb | host memory gb | route to | matched rule | default used | prompt index | outputs | decode tokens per sec |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 39.05 | 56.93 | 15 | 191 | 3 | 25.83 | 51.5 | Qwen3-1.7B-GGUF | yes | 0 | -- | ||
| 2 | 40.99 | 56.16 | 17 | 256 | 3 | 25.83 | 51.4 | Qwen3-1.7B-GGUF | yes | 1 | -- | ||
| 3 | 47.17 | 56.05 | 19 | 256 | 3 | 25.83 | 51.5 | Qwen3-1.7B-GGUF | yes | 2 | -- | ||
| 4 | 48.15 | 55.27 | 18 | 256 | 3 | 25.83 | 51.6 | Qwen3-1.7B-GGUF | yes | 3 | -- | ||
| 5 | 47.78 | 55.63 | 20 | 256 | 3 | 25.83 | 51.6 | Qwen3-1.7B-GGUF | yes | 4 | -- | ||
| 6 | 38.1 | 56.07 | 17 | 256 | 3 | 25.83 | 51.6 | Qwen3-1.7B-GGUF | yes | 5 | -- | ||
| 7 | 46.13 | 56.26 | 14 | 256 | 3 | 25.83 | 51.7 | Qwen3-1.7B-GGUF | yes | 6 | -- | ||
| 8 | 46.73 | 54.59 | 17 | 256 | 3 | 25.83 | 51.7 | Qwen3-1.7B-GGUF | yes | 7 | -- | ||
| 9 | 36.42 | 54.95 | 16 | 256 | 3 | 25.83 | 51.7 | Qwen3-1.7B-GGUF | yes | 8 | -- | ||
| 10 | 50.12 | 56.6 | 16 | 256 | 3 | 25.83 | 51.7 | Qwen3-1.7B-GGUF | yes | 9 | -- | ||
| 11 | 595.1 | 71.7 | 29 | 256 | 0 | 25.83 | 51.8 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 10 | -- | |
| 12 | 173 | 81.89 | 37 | 256 | 0 | 25.83 | 51.9 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 11 | -- | |
| 13 | 194.1 | 82.59 | 57 | 256 | 0 | 25.83 | 52.1 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 12 | -- | |
| 14 | 110.9 | 96.62 | 28 | 256 | 0 | 25.83 | 52.2 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 13 | -- | |
| 15 | 99.94 | 98.81 | 30 | 256 | 0 | 25.83 | 52.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 14 | -- | |
| 16 | 96.14 | 100.1 | 20 | 256 | 0 | 25.83 | 52.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 15 | -- | |
| 17 | 153.3 | 99.99 | 42 | 256 | 0 | 25.83 | 52.5 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 16 | -- | |
| 18 | 103.2 | 100.5 | 29 | 256 | 0 | 25.83 | 52.6 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 17 | -- | |
| 19 | 95.58 | 101 | 24 | 256 | 0 | 25.83 | 52.9 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 18 | -- | |
| 20 | 162.9 | 100.6 | 42 | 256 | 0 | 25.83 | 52.9 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 19 | -- | |
| 21 | 0 | -- | 27 | 256 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 20 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 22 | 0 | -- | 23 | 256 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 21 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 23 | 0 | -- | 22 | 256 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 22 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 24 | 0 | -- | 21 | 256 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 23 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 25 | 0 | -- | 21 | 256 | 0 | 25.83 | 52.9 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 24 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 26 | 0 | -- | 19 | 256 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 25 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 27 | 0 | -- | 19 | 256 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 26 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 28 | 0 | -- | 18 | 256 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 27 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 29 | 367.8 | 55.45 | 23 | 256 | 3 | 25.83 | 53 | Qwen3-1.7B-GGUF | yes | 28 | -- | ||
| 30 | 0 | -- | 18 | 256 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 29 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 31 | 0 | -- | 182 | 7 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 30 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 32 | 0 | -- | 178 | 115 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 31 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 33 | 0 | -- | 185 | 24 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 32 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 34 | 0 | -- | 182 | 76 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 33 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 35 | 0 | -- | 184 | 66 | 0 | 25.83 | 53 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 34 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 36 | 139.2 | 81.96 | 165 | 256 | 161 | 25.83 | 53 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 35 | -- | |
| 37 | 56.81 | 98.02 | 167 | 256 | 163 | 25.83 | 53.1 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 36 | -- | |
| 38 | 52.89 | 99.59 | 167 | 256 | 163 | 25.83 | 53.1 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 37 | -- | |
| 39 | 54.34 | 97.96 | 166 | 256 | 162 | 25.83 | 53.1 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 38 | -- | |
| 40 | 47.26 | 99.04 | 164 | 256 | 160 | 25.83 | 53.1 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 39 | -- | |
| 41 | 159.5 | 55.86 | 15 | 219 | 3 | 25.83 | 53 | Qwen3-1.7B-GGUF | yes | 0 | -- | ||
| 42 | 43.72 | 55.96 | 17 | 256 | 3 | 25.83 | 53 | Qwen3-1.7B-GGUF | yes | 1 | -- | ||
| 43 | 48.85 | 54.41 | 19 | 256 | 3 | 25.83 | 53.1 | Qwen3-1.7B-GGUF | yes | 2 | -- | ||
| 44 | 56.4 | 54.11 | 18 | 256 | 3 | 25.83 | 53.1 | Qwen3-1.7B-GGUF | yes | 3 | -- | ||
| 45 | 60.88 | 55.51 | 20 | 256 | 3 | 25.83 | 53.1 | Qwen3-1.7B-GGUF | yes | 4 | -- | ||
| 46 | 51.76 | 55.63 | 17 | 256 | 3 | 25.83 | 53.2 | Qwen3-1.7B-GGUF | yes | 5 | -- | ||
| 47 | 46.72 | 55.73 | 14 | 256 | 3 | 25.83 | 53.2 | Qwen3-1.7B-GGUF | yes | 6 | -- | ||
| 48 | 52.18 | 55.65 | 17 | 256 | 3 | 25.83 | 53.2 | Qwen3-1.7B-GGUF | yes | 7 | -- | ||
| 49 | 50.23 | 55.2 | 16 | 256 | 3 | 25.83 | 53.3 | Qwen3-1.7B-GGUF | yes | 8 | -- | ||
| 50 | 39.02 | 53.2 | 16 | 256 | 3 | 25.83 | 53.3 | Qwen3-1.7B-GGUF | yes | 9 | -- | ||
| 51 | 239.2 | 74.74 | 29 | 256 | 0 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 10 | -- | |
| 52 | 187.7 | 81.94 | 37 | 256 | 0 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 11 | -- | |
| 53 | 209.1 | 80.59 | 57 | 256 | 0 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 12 | -- | |
| 54 | 110 | 82.53 | 28 | 256 | 0 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 13 | -- | |
| 55 | 106.9 | 83.22 | 30 | 256 | 0 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 14 | -- | |
| 56 | 101.9 | 82.77 | 20 | 256 | 0 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 15 | -- | |
| 57 | 158.9 | 82.78 | 42 | 256 | 0 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 16 | -- | |
| 58 | 108.8 | 82.87 | 29 | 256 | 0 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 17 | -- | |
| 59 | 116.4 | 82.94 | 24 | 256 | 0 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 18 | -- | |
| 60 | 183.5 | 82.93 | 42 | 256 | 0 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 19 | -- | |
| 61 | 0 | -- | 27 | 256 | 0 | 25.83 | 53.5 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 20 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 62 | 0 | -- | 23 | 256 | 0 | 25.83 | 53.4 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 21 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 63 | 0 | -- | 22 | 256 | 0 | 25.83 | 53.4 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 22 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 64 | 0 | -- | 21 | 256 | 0 | 25.83 | 53.4 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 23 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 65 | 0 | -- | 21 | 256 | 0 | 25.83 | 53.4 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 24 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 66 | 0 | -- | 19 | 256 | 0 | 25.83 | 53.4 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 25 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 67 | 0 | -- | 19 | 256 | 0 | 25.83 | 53.4 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 26 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 68 | 0 | -- | 18 | 256 | 0 | 25.83 | 53.5 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 27 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 69 | 658.6 | 55.28 | 23 | 256 | 3 | 25.83 | 53.5 | Qwen3-1.7B-GGUF | yes | 28 | -- | ||
| 70 | 0 | -- | 18 | 256 | 0 | 25.83 | 53.5 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 29 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 71 | 0 | -- | 182 | 7 | 0 | 25.83 | 53.5 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 30 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 72 | 0 | -- | 178 | 128 | 0 | 25.83 | 53.5 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 31 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 73 | 0 | -- | 185 | 22 | 0 | 25.83 | 53.5 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 32 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 74 | 0 | -- | 182 | 87 | 0 | 25.83 | 53.5 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 33 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 75 | 0 | -- | 184 | 71 | 0 | 25.83 | 53.5 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 34 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 76 | 55.92 | 93.55 | 165 | 256 | 161 | 25.83 | 53.3 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 35 | -- | |
| 77 | 46.42 | 98.67 | 167 | 256 | 163 | 25.83 | 53.5 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 36 | -- | |
| 78 | 49.08 | 99.32 | 167 | 256 | 163 | 25.83 | 53.3 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 37 | -- | |
| 79 | 46.92 | 100.1 | 166 | 256 | 162 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 38 | -- | |
| 80 | 48.57 | 99.58 | 164 | 256 | 160 | 25.83 | 53.4 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 39 | -- | |
| 81 | 43.58 | 54.54 | 15 | 243 | 3 | 25.83 | 53.4 | Qwen3-1.7B-GGUF | yes | 0 | -- | ||
| 82 | 38.31 | 56 | 17 | 256 | 3 | 25.83 | 53.6 | Qwen3-1.7B-GGUF | yes | 1 | -- | ||
| 83 | 55.33 | 54.6 | 19 | 227 | 3 | 25.83 | 53.4 | Qwen3-1.7B-GGUF | yes | 2 | -- | ||
| 84 | 46.82 | 54.84 | 18 | 256 | 3 | 25.83 | 53.5 | Qwen3-1.7B-GGUF | yes | 3 | -- | ||
| 85 | 58.94 | 54.57 | 20 | 256 | 3 | 25.83 | 53.5 | Qwen3-1.7B-GGUF | yes | 4 | -- | ||
| 86 | 59.2 | 53.91 | 17 | 256 | 3 | 25.83 | 53.5 | Qwen3-1.7B-GGUF | yes | 5 | -- | ||
| 87 | 53.27 | 54.81 | 14 | 256 | 3 | 25.83 | 53.6 | Qwen3-1.7B-GGUF | yes | 6 | -- | ||
| 88 | 61.42 | 53.93 | 17 | 256 | 3 | 25.83 | 53.6 | Qwen3-1.7B-GGUF | yes | 7 | -- | ||
| 89 | 46.89 | 55.51 | 16 | 256 | 3 | 25.83 | 53.6 | Qwen3-1.7B-GGUF | yes | 8 | -- | ||
| 90 | 43.25 | 54.96 | 16 | 256 | 3 | 25.83 | 53.7 | Qwen3-1.7B-GGUF | yes | 9 | -- | ||
| 91 | 132 | 80 | 29 | 256 | 0 | 25.83 | 54 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 10 | -- | |
| 92 | 188.8 | 82.44 | 37 | 256 | 0 | 25.83 | 53.8 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 11 | -- | |
| 93 | 191.9 | 82.62 | 57 | 256 | 0 | 25.83 | 53.8 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 12 | -- | |
| 94 | 109.2 | 82.77 | 28 | 256 | 0 | 25.83 | 53.8 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 13 | -- | |
| 95 | 107.7 | 82.67 | 30 | 256 | 0 | 25.83 | 53.9 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 14 | -- | |
| 96 | 109.4 | 81.12 | 20 | 256 | 0 | 25.83 | 53.8 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 15 | -- | |
| 97 | 163.2 | 82.8 | 42 | 256 | 0 | 25.83 | 53.8 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 16 | -- | |
| 98 | 110.2 | 83.23 | 29 | 256 | 0 | 25.83 | 53.8 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 17 | -- | |
| 99 | 118.1 | 83.4 | 24 | 256 | 0 | 25.83 | 53.8 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 18 | -- | |
| 100 | 175 | 83.38 | 42 | 256 | 0 | 25.83 | 53.9 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 19 | -- | |
| 101 | 0 | -- | 27 | 256 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 20 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 102 | 0 | -- | 23 | 256 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 21 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 103 | 0 | -- | 22 | 256 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 22 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 104 | 0 | -- | 21 | 256 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 23 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 105 | 0 | -- | 21 | 256 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 24 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 106 | 0 | -- | 19 | 256 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 25 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 107 | 0 | -- | 19 | 256 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 26 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 108 | 0 | -- | 18 | 256 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 27 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 109 | 66.5 | 54.13 | 23 | 256 | 3 | 25.83 | 53.8 | Qwen3-1.7B-GGUF | yes | 28 | -- | ||
| 110 | 0 | -- | 18 | 256 | 0 | 25.83 | 54 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 29 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 111 | 0 | -- | 182 | 42 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 30 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 112 | 0 | -- | 178 | 120 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 31 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 113 | 0 | -- | 185 | 50 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 32 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 114 | 0 | -- | 182 | 73 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 33 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 115 | 0 | -- | 184 | 70 | 0 | 25.83 | 53.8 | devup.deepseek-ai/DeepSeek-V4-Pro | analysis-or-long-to-cloud | no | 34 | {"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}} | |
| 116 | 62.84 | 78.29 | 165 | 256 | 161 | 25.83 | 53.7 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 35 | -- | |
| 117 | 56.35 | 82.22 | 167 | 256 | 163 | 25.83 | 53.7 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 36 | -- | |
| 118 | 47.4 | 90.42 | 167 | 256 | 163 | 25.83 | 53.7 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 37 | -- | |
| 119 | 51.28 | 97.96 | 166 | 256 | 162 | 25.83 | 53.8 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 38 | -- | |
| 120 | 48.56 | 99.48 | 164 | 256 | 160 | 25.83 | 53.8 | Qwen3.5-35B-A3B-GGUF | code-to-big-local | no | 39 | -- |
Model and configuration
| model.name | user.LM-D-three-tier |
| model.class | router |
| model.quantization | none |
| model.context_length | 4096 |
| config.compute_engine | router |
| config.backend | collection.router |
| config.os | linux |
| config.power_state | plugged |
| config.router_default_model | Qwen3-1.7B-GGUF |
Environment
| cpu | AMD Ryzen 9 9950X3D 16-Core Processor |
| device_model | AMD Ryzen 9 9950X3D 16-Core Processor |
| dgpu | AMD Radeon AI PRO R9700 (gfx1201) |
| dgpu_family | gfx1201 |
| driver_version | unknown |
| igpu | AMD Ryzen 9 9950X3D 16-Core Processor (gfx1036) |
| igpu_family | gfx1036 |
| lemonade_version | 2026.46.2 |
| memory_gb | 59.46 |
| os_version | Linux-7.0.0-42-generic (Ubuntu 26.04) |
All metrics
| prefill.tokens_per_sec | 495.7 |
| prefill.stddev | 1,145.6 |
| prefill.p95 | 3,412.8 |
| prefill.ci95_half_width | 254.2 |
| decode.tokens_per_sec | 74.54 |
| decode.inter_token_latency_ms | 13.42 |
| decode.stddev | 17.96 |
| decode.p95 | 100.1 |
| decode.ci95_half_width | 3.985 |
| ttft_ms | 105.5 |
| trial_count | 120 |
| peak_memory_gb | 25.83 |
| prompt_sha256 | 5061664e3d07a0e9e1e54868eddcbf4da9b5ccbf0651c94994b016bf595d30a8 |
| routing.requests | 120 |
| routing.candidates.Qwen3-1.7B-GGUF.requests | 33 |
| routing.candidates.Qwen3-1.7B-GGUF.prompt_tokens | 576 |
| routing.candidates.Qwen3-1.7B-GGUF.output_tokens | 8304 |
| routing.candidates.Qwen3-1.7B-GGUF.cost_usd | 0 |
| routing.candidates.Qwen3-1.7B-GGUF.price_source | local_default |
| routing.candidates.Qwen3-1.7B-GGUF.share | 0.275 |
| routing.candidates.Qwen3.5-35B-A3B-GGUF.requests | 45 |
| routing.candidates.Qwen3.5-35B-A3B-GGUF.prompt_tokens | 3501 |
| routing.candidates.Qwen3.5-35B-A3B-GGUF.output_tokens | 11520 |
| routing.candidates.Qwen3.5-35B-A3B-GGUF.cost_usd | 0 |
| routing.candidates.Qwen3.5-35B-A3B-GGUF.price_source | local_default |
| routing.candidates.Qwen3.5-35B-A3B-GGUF.share | 0.375 |
| routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.requests | 42 |
| routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.prompt_tokens | 3297 |
| routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.output_tokens | 7870 |
| routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.cost_usd | 0.01477 |
| routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.price_source | override |
| routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.share | 0.35 |
| routing.local_share | 0.65 |
| routing.free_share | 0.65 |
| routing.routed_cost_usd | 0.01477 |
| routing.unpriced_requests | 0 |
| routing.prompt_set_size | 40 |
| routing.baseline_model | devup.deepseek-ai/DeepSeek-V4-Pro |
| routing.baseline_method | measured |
| routing.baseline_estimated_requests | 0 |
| routing.baseline_cost_usd | 0.04061 |
| routing.savings_usd | 0.02584 |
| routing.savings_pct | 63.63 |
| ttft_ms_stddev | 104.7 |
| ttft_ms_ci95_half_width | 23.23 |