Reasoning Gauntlet — reasoning-gauntlet-v1
Code Generation Tool Calling RAG / Q&A Reasoning / Math | About

The Gauntlet

Six escalating tests, run identically across every model. Find what actually holds up on your hardware — organized by VRAM tier, not marketing size class.

Sub-4GB

6 models tested
1.8 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 149.0 tok · 2.27s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 895.8 tok · 4.73s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 591.2 tok · 3.26s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 386.0 tok · 2.24s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 579.2 tok · 3.2s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 5/5 · 269.2 tok · 1.66s
30/30
3.4 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 272.0 tok · 3.77s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 1018.0 tok · 7.54s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 575.4 tok · 4.43s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 324.0 tok · 2.66s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 562.0 tok · 4.31s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 5/5 · 585.6 tok · 4.47s
30/30
llama3.2:3b Champion
3.1 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 115.0 tok · 1.91s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 152.0 tok · 0.72s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 238.0 tok · 1.08s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 135.8 tok · 0.7s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 249.2 tok · 1.11s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 5/5 · 152.0 tok · 0.75s
30/30
2.7 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 427.8 tok · 2.22s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 1149.2 tok · 4.82s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 852.2 tok · 3.61s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 822.0 tok · 3.46s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 1328.0 tok · 5.57s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 1/5 · 974.6 tok · 4.17s
26/30
2.9 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 65.0 tok · 0.71s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 175.8 tok · 0.91s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 0/5 · 24.0 tok · 0.21s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 175.8 tok · 0.91s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 275.0 tok · 1.38s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 0/5 · 115.2 tok · 0.62s
20/30
3.7 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 46.0 tok · 0.81s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 0/5 · 232.2 tok · 1.26s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 137.0 tok · 0.88s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 113.0 tok · 0.69s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 169.4 tok · 0.95s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 0/5 · 113.0 tok · 0.68s
20/30

4-8GB

4 models tested
qwen3:8b Champion
6.3 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 383.6 tok · 3.53s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 1356.8 tok · 9.63s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 779.0 tok · 5.56s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 1272.0 tok · 9.04s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 2112.2 tok · 15.03s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 5/5 · 1017.2 tok · 7.23s
30/30
5.7 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 1850.0 tok · 16.07s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 0/5 · 4096.0 tok · 32.79s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 1007.0 tok · 8.22s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 1037.0 tok · 8.63s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 1611.0 tok · 13.11s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 5/5 · 945.0 tok · 7.83s
25/30
5.6 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 78.0 tok · 0.99s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 370.0 tok · 2.34s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 193.0 tok · 1.23s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 224.0 tok · 1.43s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 310.0 tok · 1.95s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 0/5 · 143.0 tok · 0.93s
25/30
5.1 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 80.0 tok · 1.13s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 465.0 tok · 3.16s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 156.6 tok · 1.17s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 10.0 tok · 0.2s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 437.4 tok · 2.98s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 0/5 · 381.0 tok · 2.64s
25/30

8-12GB

4 models tested
8.0 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 299.0 tok · 4.83s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 908.4 tok · 10.56s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 522.2 tok · 6.38s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 445.4 tok · 5.5s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 550.6 tok · 6.66s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 5/5 · 474.4 tok · 5.82s
30/30
8.4 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 295.4 tok · 4.43s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 1192.2 tok · 14.65s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 518.6 tok · 6.69s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 297.8 tok · 3.97s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 635.0 tok · 10.04s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 5/5 · 503.0 tok · 6.37s
30/30
deepseek-r1:14b Champion
10 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 237.0 tok · 5.02s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 513.6 tok · 6.29s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 877.0 tok · 10.71s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 430.0 tok · 5.42s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 420.8 tok · 5.34s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 5/5 · 472.8 tok · 6.0s
30/30
10 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 327.0 tok · 4.91s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 1890.4 tok · 22.69s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 911.0 tok · 11.0s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 1375.0 tok · 16.54s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 2202.2 tok · 26.6s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 5/5 · 2088.6 tok · 25.14s
30/30

12GB+

3 models tested
15 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 78.0 tok · 3.62s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 264.0 tok · 5.0s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 156.0 tok · 2.9s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 165.0 tok · 3.05s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 201.6 tok · 3.72s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 5/5 · 183.8 tok · 3.4s
30/30
17 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 298.0 tok · 3.95s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 1254.2 tok · 9.96s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 530.4 tok · 4.5s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 532.0 tok · 4.52s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 754.6 tok · 6.3s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 5/5 · 451.0 tok · 3.84s
30/30
22 GB measured VRAM
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 1 — floor: 5/5 · 331.8 tok · 10.76s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 2 — multi-step: 5/5 · 1417.6 tok · 35.75s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 3 — logical deduction: 5/5 · 688.6 tok · 17.35s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 4 — adversarial (CRT trap): 5/5 · 923.2 tok · 23.05s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 5 — combinatorics: 5/5 · 1371.2 tok · 34.17s
[This tool call rendered an interactive widget in the chat. The user can already see the result — do not repeat it in text or with another visualization tool.] Tier 6 — distractor filtering: 2/5 · 2806.0 tok · 76.01s
27/30