Skip to content
VECTOR WIREAI INTELLIGENCE
⌘K
UTC
Models743Deals30d268Policy30d68, 11 pendingSources1,019
Pro Mode — raw scores, one benchmark under the lensofficial card harness
The field · 31 measured
3.3398

31 models measured, most on the official card harness. GPT-5.2 tops the board at 98.

31 measured·Lifecycle Updated Oct 9, 2026Score basis
#◆Vendor$/M in$/M out
Leader — ignites whiteField — fill cools with scoreBars show 3.33–98 · ◆ solid = first-party or better · outlined = arm's-length
01GPT-5.2OpenAI98$1.8/M$14.0/M$0.080
02Gemini 3 ProGoogle96$2.0/M$12.0/M$0.073
03Claude Opus 4.6Anthropic95.6$5.0/M$25.0/M$0.157
04GPT-5OpenAI94.6$1.3/M$10.0/M$0.059
05Kimi K2 ThinkingMoonshot94.5$0.60/M$2.5/M$0.016
06Nemotron 3 Super 120B A12BNVIDIA92.23P$0.30/M$0.90/M$0.007
07Grok 4SpaceXAI91.73P$3.0/M$15.0/M$0.098
08Claude Opus 4.5Anthropic91$5.0/M$25.0/M$0.165
09NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-BF16NVIDIA89.73P———
10DeepSeek-V3.2DeepSeek89.33P$0.28/M$0.42/M$0.004
11Granite 4.2 30BIBM89.17$0.16/M$0.65/M$0.005
12Claude Sonnet 4.5+1 altAnthropic88$3.0/M$15.0/M$0.102
13Granite 4.2 8B+1 altIBM86.67$0.06/M$0.25/M$0.002
14MiniMax M2.5MiniMax86.3$0.27/M$1.1/M$0.008
15MiniMax M2.1MiniMax83$0.30/M$1.2/M$0.009
16Granite 4.2 3B+1 altIBM78.33$0.03/M$0.12/M$0.001
17Qwen3 30B A3B 2507 ThinkingAlibaba71.673P$0.20/M$2.4/M$0.018
18Qwen3.5 9B+1 altAlibaba56.07$0.14/M$0.20/M$0.003
19LFM2.5-2.6B+1 altLiquid AI51.87———
20Kimi K2 InstructMoonshot513P$0.57/M$2.3/M$0.028
21Qwen3.5 4B+2 altsAlibaba49.33$0.03/M$0.15/M$0.002
22LFM2.5-8B-A1BLiquid AI42.533P———
23Qwen3 1.7BAlibaba36.273P———
24Gemma 4 E4B+2 altsGoogle34.27$0.02/M$0.10/M$0.002
25LFM2.5-1.2B-ThinkingLiquid AI31.733P———
26Gemma 4 E2B+2 altsGoogle26.33———
27LFM2-8B-A1B (Non-Reasoning)Liquid AI203P———
28LFM2.5-1.2B-Instruct+1 altLiquid AI143P———
29Qwen3 1.7B (instruct)+1 altAlibaba9.333P———
30Granite 4.0 H TinyIBM4.933P———
31Granite 4.0 1B+1 altIBM3.333P———
◆ solid = independently verified · aggregator attested · vendor attributed — outlined = cross-reference · source-attributed. Dimmed ◆ = confidence inferred. “+N alts” = alternate disclosures in the drawer with why-demoted. Prices per million tokens; $/point = input price ÷ score. Every row opens its model dossier.

Verification: 31 models scored · 0 independently verified · 14 vendor cross-reference · 17 vendor-reported · 0 with source disagreement. How these tiers are assigned