← Leaderboard Router run · 8904cd59-1f61-42b5-a980-38f53c8d59e3

user.LM-D-three-tier

AMD Radeon AI PRO R9700 (gfx1201) · profile Radeon-AI-PRO-R9700

Routerrouterroutercollection.routervalid 2026-11-14 15:42 UTC
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.

Trials

Each measured trial after warm-up, from the run's trial sidecar. The summary above is computed from these.

#ttft msdecode tokens per secprompt tokensoutput tokenscache tokensvram gbhost memory gbroute tomatched ruledefault usedprompt indexoutputsdecode tokens per sec
139.0556.9315191325.8351.5Qwen3-1.7B-GGUFyes0--
240.9956.1617256325.8351.4Qwen3-1.7B-GGUFyes1--
347.1756.0519256325.8351.5Qwen3-1.7B-GGUFyes2--
448.1555.2718256325.8351.6Qwen3-1.7B-GGUFyes3--
547.7855.6320256325.8351.6Qwen3-1.7B-GGUFyes4--
638.156.0717256325.8351.6Qwen3-1.7B-GGUFyes5--
746.1356.2614256325.8351.7Qwen3-1.7B-GGUFyes6--
846.7354.5917256325.8351.7Qwen3-1.7B-GGUFyes7--
936.4254.9516256325.8351.7Qwen3-1.7B-GGUFyes8--
1050.1256.616256325.8351.7Qwen3-1.7B-GGUFyes9--
11595.171.729256025.8351.8Qwen3.5-35B-A3B-GGUFcode-to-big-localno10--
1217381.8937256025.8351.9Qwen3.5-35B-A3B-GGUFcode-to-big-localno11--
13194.182.5957256025.8352.1Qwen3.5-35B-A3B-GGUFcode-to-big-localno12--
14110.996.6228256025.8352.2Qwen3.5-35B-A3B-GGUFcode-to-big-localno13--
1599.9498.8130256025.8352.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno14--
1696.14100.120256025.8352.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno15--
17153.399.9942256025.8352.5Qwen3.5-35B-A3B-GGUFcode-to-big-localno16--
18103.2100.529256025.8352.6Qwen3.5-35B-A3B-GGUFcode-to-big-localno17--
1995.5810124256025.8352.9Qwen3.5-35B-A3B-GGUFcode-to-big-localno18--
20162.9100.642256025.8352.9Qwen3.5-35B-A3B-GGUFcode-to-big-localno19--
210--27256025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno20{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
220--23256025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno21{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
230--22256025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno22{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
240--21256025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno23{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
250--21256025.8352.9devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno24{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
260--19256025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno25{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
270--19256025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno26{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
280--18256025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno27{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
29367.855.4523256325.8353Qwen3-1.7B-GGUFyes28--
300--18256025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno29{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
310--1827025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno30{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
320--178115025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno31{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
330--18524025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno32{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
340--18276025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno33{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
350--18466025.8353devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno34{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
36139.281.9616525616125.8353Qwen3.5-35B-A3B-GGUFcode-to-big-localno35--
3756.8198.0216725616325.8353.1Qwen3.5-35B-A3B-GGUFcode-to-big-localno36--
3852.8999.5916725616325.8353.1Qwen3.5-35B-A3B-GGUFcode-to-big-localno37--
3954.3497.9616625616225.8353.1Qwen3.5-35B-A3B-GGUFcode-to-big-localno38--
4047.2699.0416425616025.8353.1Qwen3.5-35B-A3B-GGUFcode-to-big-localno39--
41159.555.8615219325.8353Qwen3-1.7B-GGUFyes0--
4243.7255.9617256325.8353Qwen3-1.7B-GGUFyes1--
4348.8554.4119256325.8353.1Qwen3-1.7B-GGUFyes2--
4456.454.1118256325.8353.1Qwen3-1.7B-GGUFyes3--
4560.8855.5120256325.8353.1Qwen3-1.7B-GGUFyes4--
4651.7655.6317256325.8353.2Qwen3-1.7B-GGUFyes5--
4746.7255.7314256325.8353.2Qwen3-1.7B-GGUFyes6--
4852.1855.6517256325.8353.2Qwen3-1.7B-GGUFyes7--
4950.2355.216256325.8353.3Qwen3-1.7B-GGUFyes8--
5039.0253.216256325.8353.3Qwen3-1.7B-GGUFyes9--
51239.274.7429256025.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno10--
52187.781.9437256025.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno11--
53209.180.5957256025.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno12--
5411082.5328256025.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno13--
55106.983.2230256025.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno14--
56101.982.7720256025.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno15--
57158.982.7842256025.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno16--
58108.882.8729256025.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno17--
59116.482.9424256025.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno18--
60183.582.9342256025.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno19--
610--27256025.8353.5devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno20{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
620--23256025.8353.4devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno21{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
630--22256025.8353.4devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno22{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
640--21256025.8353.4devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno23{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
650--21256025.8353.4devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno24{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
660--19256025.8353.4devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno25{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
670--19256025.8353.4devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno26{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
680--18256025.8353.5devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno27{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
69658.655.2823256325.8353.5Qwen3-1.7B-GGUFyes28--
700--18256025.8353.5devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno29{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
710--1827025.8353.5devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno30{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
720--178128025.8353.5devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno31{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
730--18522025.8353.5devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno32{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
740--18287025.8353.5devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno33{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
750--18471025.8353.5devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno34{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
7655.9293.5516525616125.8353.3Qwen3.5-35B-A3B-GGUFcode-to-big-localno35--
7746.4298.6716725616325.8353.5Qwen3.5-35B-A3B-GGUFcode-to-big-localno36--
7849.0899.3216725616325.8353.3Qwen3.5-35B-A3B-GGUFcode-to-big-localno37--
7946.92100.116625616225.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno38--
8048.5799.5816425616025.8353.4Qwen3.5-35B-A3B-GGUFcode-to-big-localno39--
8143.5854.5415243325.8353.4Qwen3-1.7B-GGUFyes0--
8238.315617256325.8353.6Qwen3-1.7B-GGUFyes1--
8355.3354.619227325.8353.4Qwen3-1.7B-GGUFyes2--
8446.8254.8418256325.8353.5Qwen3-1.7B-GGUFyes3--
8558.9454.5720256325.8353.5Qwen3-1.7B-GGUFyes4--
8659.253.9117256325.8353.5Qwen3-1.7B-GGUFyes5--
8753.2754.8114256325.8353.6Qwen3-1.7B-GGUFyes6--
8861.4253.9317256325.8353.6Qwen3-1.7B-GGUFyes7--
8946.8955.5116256325.8353.6Qwen3-1.7B-GGUFyes8--
9043.2554.9616256325.8353.7Qwen3-1.7B-GGUFyes9--
911328029256025.8354Qwen3.5-35B-A3B-GGUFcode-to-big-localno10--
92188.882.4437256025.8353.8Qwen3.5-35B-A3B-GGUFcode-to-big-localno11--
93191.982.6257256025.8353.8Qwen3.5-35B-A3B-GGUFcode-to-big-localno12--
94109.282.7728256025.8353.8Qwen3.5-35B-A3B-GGUFcode-to-big-localno13--
95107.782.6730256025.8353.9Qwen3.5-35B-A3B-GGUFcode-to-big-localno14--
96109.481.1220256025.8353.8Qwen3.5-35B-A3B-GGUFcode-to-big-localno15--
97163.282.842256025.8353.8Qwen3.5-35B-A3B-GGUFcode-to-big-localno16--
98110.283.2329256025.8353.8Qwen3.5-35B-A3B-GGUFcode-to-big-localno17--
99118.183.424256025.8353.8Qwen3.5-35B-A3B-GGUFcode-to-big-localno18--
10017583.3842256025.8353.9Qwen3.5-35B-A3B-GGUFcode-to-big-localno19--
1010--27256025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno20{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1020--23256025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno21{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1030--22256025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno22{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1040--21256025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno23{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1050--21256025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno24{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1060--19256025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno25{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1070--19256025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno26{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1080--18256025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno27{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
10966.554.1323256325.8353.8Qwen3-1.7B-GGUFyes28--
1100--18256025.8354devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno29{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1110--18242025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno30{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1120--178120025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno31{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1130--18550025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno32{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1140--18273025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno33{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
1150--18470025.8353.8devup.deepseek-ai/DeepSeek-V4-Proanalysis-or-long-to-cloudno34{"estimated_cost": {"cost_input_per_million": 455.0, "cost_output_per_million": 910.0}}
11662.8478.2916525616125.8353.7Qwen3.5-35B-A3B-GGUFcode-to-big-localno35--
11756.3582.2216725616325.8353.7Qwen3.5-35B-A3B-GGUFcode-to-big-localno36--
11847.490.4216725616325.8353.7Qwen3.5-35B-A3B-GGUFcode-to-big-localno37--
11951.2897.9616625616225.8353.8Qwen3.5-35B-A3B-GGUFcode-to-big-localno38--
12048.5699.4816425616025.8353.8Qwen3.5-35B-A3B-GGUFcode-to-big-localno39--

Model and configuration

model.nameuser.LM-D-three-tier
model.classrouter
model.quantizationnone
model.context_length4096
config.compute_enginerouter
config.backendcollection.router
config.oslinux
config.power_stateplugged
config.router_default_modelQwen3-1.7B-GGUF

Environment

cpuAMD Ryzen 9 9950X3D 16-Core Processor
device_modelAMD Ryzen 9 9950X3D 16-Core Processor
dgpuAMD Radeon AI PRO R9700 (gfx1201)
dgpu_familygfx1201
driver_versionunknown
igpuAMD Ryzen 9 9950X3D 16-Core Processor (gfx1036)
igpu_familygfx1036
lemonade_version2026.46.2
memory_gb59.46
os_versionLinux-7.0.0-42-generic (Ubuntu 26.04)

All metrics

prefill.tokens_per_sec495.7
prefill.stddev1,145.6
prefill.p953,412.8
prefill.ci95_half_width254.2
decode.tokens_per_sec74.54
decode.inter_token_latency_ms13.42
decode.stddev17.96
decode.p95100.1
decode.ci95_half_width3.985
ttft_ms105.5
trial_count120
peak_memory_gb25.83
prompt_sha2565061664e3d07a0e9e1e54868eddcbf4da9b5ccbf0651c94994b016bf595d30a8
routing.requests120
routing.candidates.Qwen3-1.7B-GGUF.requests33
routing.candidates.Qwen3-1.7B-GGUF.prompt_tokens576
routing.candidates.Qwen3-1.7B-GGUF.output_tokens8304
routing.candidates.Qwen3-1.7B-GGUF.cost_usd0
routing.candidates.Qwen3-1.7B-GGUF.price_sourcelocal_default
routing.candidates.Qwen3-1.7B-GGUF.share0.275
routing.candidates.Qwen3.5-35B-A3B-GGUF.requests45
routing.candidates.Qwen3.5-35B-A3B-GGUF.prompt_tokens3501
routing.candidates.Qwen3.5-35B-A3B-GGUF.output_tokens11520
routing.candidates.Qwen3.5-35B-A3B-GGUF.cost_usd0
routing.candidates.Qwen3.5-35B-A3B-GGUF.price_sourcelocal_default
routing.candidates.Qwen3.5-35B-A3B-GGUF.share0.375
routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.requests42
routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.prompt_tokens3297
routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.output_tokens7870
routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.cost_usd0.01477
routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.price_sourceoverride
routing.candidates.devup.deepseek-ai/DeepSeek-V4-Pro.share0.35
routing.local_share0.65
routing.free_share0.65
routing.routed_cost_usd0.01477
routing.unpriced_requests0
routing.prompt_set_size40
routing.baseline_modeldevup.deepseek-ai/DeepSeek-V4-Pro
routing.baseline_methodmeasured
routing.baseline_estimated_requests0
routing.baseline_cost_usd0.04061
routing.savings_usd0.02584
routing.savings_pct63.63
ttft_ms_stddev104.7
ttft_ms_ci95_half_width23.23