Skip to content

Benchmarks

Local LLM speed results across models, backends, hardware, and power profiles. Decode tok/s is the headline metric; latency, raw engine runs, and workload context stay visible in their own views.

1181 source rows225 matching source rowslatest run May 21, 2026schemas v1-v4source content/benchmarks/runs/
Filters
Advanced filters

Full row-level explorer. This is the place for raw shapes, hardware probes, cache/ppfix rows, dense power caps, and reruns.

1.2BQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselinechat1
211.5
18ms4.7
1.2BQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselinecodegen1
210.6
24ms4.7
1.2BQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselinerag1
207.6
16ms4.8
1.2BQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselineagent1
199.2
77ms5.0
8B-A1BQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselinechat1
154.8
49ms6.5
8B-A1BQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselinecodegen1
153.8
75ms6.5
8B-A1BQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselinerag1
152.6
25ms6.6
8B-A1BQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselineagent1
149.2
21ms6.7
1.2BQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselineagent4
121.7
404ms8.2
E2B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselinechat1
89.4
87ms11.2
E2B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselinecodegen1
88.1
103ms11.3
E2B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselinerag1
87.0
364ms11.5
E2B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselineagent1
85.7
101ms11.7
4b-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinechat1
66.3
59ms15.1
4b-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinecodegen1
65.0
99ms15.4
4b-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent1
64.5
426ms15.5
4b-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinerag1
63.7
325ms15.7
8B-A1BQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselineagent4
63.6
506ms15.7
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselinechat1
55.3
148ms18.1
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselinecodegen1
54.1
170ms18.5
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinechat1
53.8
141ms18.6
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselineagent1
53.7
446ms18.6
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselinerag1
53.2
347ms18.8
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinecodegen1
52.9
164ms18.9
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent1
52.4
561ms19.1
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinerag1
52.1
382ms19.2
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselinechat1
52.0
244ms19.2
35B-A3BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinechat1
48.9
149ms20.4
35B-A3BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinecodegen1
48.9
197ms20.4
35B-A3BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent1
48.9
639ms20.5
35B-A3BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinerag1
48.8
631ms20.5
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselinecodegen1
48.3
296ms20.7
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinechat1
47.7
209ms20.9
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselineagent1
47.3
712ms21.1
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselinerag1
47.2
590ms21.2
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinecodegen1
46.5
263ms21.5
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent1
46.0
830ms21.7
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinerag1
45.1
626ms22.2
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselineagent4
35.1
5.18s28.5
E2B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (rocm)baselineagent4
33.3
2.06s30.1
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselineagent4
24.5
7.29s40.8
4b-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent4
22.2
3.36s45.0
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent4
21.3
3.28s47.0
35B-A3BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent4
18.9
1.29s53.0
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent4
14.6
5.18s68.5
27BthinkQ3_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinechat1
14.4
333ms69.4
27BthinkQ3_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinecodegen1
14.3
413ms69.7
27BthinkQ3_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselineagent4
14.3
52.94s69.9
27BthinkQ3_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinerag1
14.3
1.53s70.0
27BthinkQ3_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselineagent1
14.3
277ms70.0
27BthinkQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinechat1
12.1
352ms82.4
27BthinkQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinecodegen1
12.1
435ms82.7
27BthinkQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselineagent1
12.1
289ms82.8
27BthinkQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselineagent4
12.1
62.69s82.8
27BthinkQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinerag1
12.1
1.53s82.9
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselinechat1
12.0
360ms83.2
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinechat1
12.0
330ms83.5
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinecodegen1
12.0
417ms83.5
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselinecodegen1
12.0
481ms83.6
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinerag1
12.0
1.73s83.6
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent1
12.0
1.75s83.6
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselinerag1
11.9
1.95s83.8
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselineagent1
11.8
1.98s84.8
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinechat1
11.6
351ms86.3
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinecodegen1
11.5
426ms86.7
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent1
11.5
1.78s87.0
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinerag1
11.5
1.73s87.1
27BthinkQ5_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinechat1
10.7
358ms93.5
27BthinkQ5_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinecodegen1
10.7
440ms93.8
27BthinkQ5_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinerag1
10.7
1.61s93.8
27BthinkQ5_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselineagent4
10.7
70.91s93.8
27BthinkQ5_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselineagent1
10.7
294ms93.8
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (cpu)baselinechat1
10.6
844ms94.3
31B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinechat1
10.6
740ms94.4
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (cpu)baselinecodegen1
10.3
1.19s96.6
31B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinecodegen1
10.3
989ms97.0
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (cpu)baselineagent1
10.3
6.49s97.5
31B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent1
10.2
3.26s98.5
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (cpu)baselinerag1
10.1
4.16s99.3
31B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinerag1
10.0
2.34s99.5
27BthinkQ6_Klegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinechat1
9.4
387ms106.3
27BthinkQ6_Klegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinecodegen1
9.4
486ms106.5
27BthinkQ6_Klegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselineagent4
9.4
80.46s106.5
27BthinkQ6_Klegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinerag1
9.4
1.66s106.5
27BthinkQ6_Klegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselineagent1
9.4
281ms106.6
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (cpu)baselinechat1
9.1
1.41s110.4
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (cpu)baselinecodegen1
8.7
1.89s115.5
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (cpu)baselineagent1
8.5
9.60s118.0
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (cpu)baselinerag1
8.3
6.29s120.6
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (vulkan)baselineagent4
7.8
3.82s128.2
27BthinkQ8_0legacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinechat1
7.7
453ms129.5
27BthinkQ8_0legacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinecodegen1
7.7
439ms129.7
27BthinkQ8_0legacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselinerag1
7.7
1.54s129.8
27BthinkQ8_0legacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselineagent4
7.7
97.84s129.8
27BthinkQ8_0legacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unifieddrv 7
llama.cpp rocm-4f13cb7 (rocm)baselineagent1
7.7
281ms129.9
E4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (cpu)baselineagent4
6.7
8.71s149.9
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent4
5.6
3.44s178.9
27BthinkQ4_K_XLlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent4
5.6
2.12s178.9
26B-A4B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b8940 (cpu)baselineagent4
3.3
13.28s298.7
31B-itQ4_K_Mlegacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent4
3.3
14.37s306.3
Decode tok/s
Headline speed metric
TTFT / TPOT
Latency context
Raw vs workload
Separate comparison contracts
Notes badge key
hardware comparable

Use these rows for GPU-to-GPU comparisons when the model, quant, backend, driver family, power policy, and benchmark shape match closely.

stack comparable

Use these rows to compare a similar software stack. They are useful, but backend, server path, driver, cache, or power settings may still influence the number.

stack realistic

Treat these as real workload measurements, not pure hardware rankings. They include prompt mix, API/server overhead, cache behavior, and local software details.

legacyOlder workload harness row.
350 W capRecorded GPU power limit.
drv 590GPU driver branch.
reasoningReasoning-token model.
Metric guide
Decode tok/s - Generation rate. Raw rows come from the engine benchmark; API rows use token intervals when available.
TTFT - Time to first token. This includes prompt processing and server/API overhead.
TPOT / ITL - Time per output token after the first token. Lower is better.
Raw Engine - llama-bench style cases intended for hardware-normalized comparison across rigs.
Workload / API - Stack-realistic measurements that include backend, server, cache, driver, and prompt behavior.
Power badges - A cap badge shows the recorded power limit. The row metadata records the cap relative to the recorded max.