Skip to content
VECTOR WIREAI INTELLIGENCE
⌘K
UTC
Models741Deals30d263Policy30d65, 14 pendingSources1,013
Pro Mode — raw scores, one benchmark under the lensofficial card harness
The field · 32 measured
46.470

32 models measured, most on the official card harness. Qwen3.5 27B tops the board at 70.

32 measured·Lifecycle Updated Oct 8, 2026Score basis
#◆Vendor$/M in$/M out
Leader — ignites whiteField — fill cools with scoreBars show 46.4–70 · ◆ solid = first-party or better · outlined = arm's-length
01Qwen3.5 27B+1 altAlibaba70$0.30/M$2.4/M$0.019
02Qwen3.6 35B A3B+1 altAlibaba69.8$0.38/M$2.3/M$0.019
03Qwen3.5 35B A3B+1 altAlibaba67.9$0.25/M$2.0/M$0.017
04Qwen3.5 122B A10BAlibaba67.6$0.40/M$3.2/M$0.027
05Gemma 4 31BGoogle67.4$0.17/M$0.40/M$0.004
06Qwen3 VL 32B ReasoningAlibaba67.4$0.16/M$0.64/M$0.006
07Qwen3 VL 2B Instruct+1 altAlibaba67.3———
08Qwen3 VL 235B A22B ReasoningAlibaba66.7$0.40/M$4.0/M$0.033
09Gemma 4 26B A4BGoogle66.1$0.07/M$0.34/M$0.003
10Qwen3 VL 30B A3B ReasoningAlibaba66$0.20/M$2.4/M$0.020
11Qwen3.5 2B+2 altsAlibaba65.5———
12Qwen3 VL Thinking (8B)Alibaba65.4$0.18/M$2.1/M$0.017
13Ministral 3 3B+1 altMistral65.2$0.10/M$0.10/M$0.002
14Qwen3 VL 4B (Reasoning)Alibaba64.1———
15Seed 1.8ByteDance63.9$0.25/M$2.0/M$0.018
16Qwen3 VL 32B InstructAlibaba63.8$0.16/M$0.64/M$0.006
17Qwen3 VL 235B A22B InstructAlibaba63.2$0.40/M$1.6/M$0.016
18Qwen3 VL 30B A3B InstructAlibaba61.5$0.20/M$0.80/M$0.008
19North-Micro-Vision-Instruct+2 altsCohere61.5———
20Qwen3 VL 8B InstructAlibaba61.1$0.18/M$0.70/M$0.007
21LFM2.5-VL-1.6B+1 altLiquid AI60.1———
22Claude Sonnet 4.5Anthropic59.9$3.0/M$15.0/M$0.150
23Gemma 4 E2B+2 altsGoogle59.8———
24Phi 3.5 Vision Instruct+1 altMicrosoft58.5———
25Qwen3 VL 4B InstructAlibaba57.6———
26Qwen2.5 VL 72B+1 altAlibaba55.16———
27InternVL3_5-4BOpenGVLab52.1———
28Qwen3.5 4BAlibaba51.7$0.03/M$0.15/M$0.002
29Gemma 4 E4BGoogle49.8$0.02/M$0.10/M$0.001
30InternVL3_5-2BOpenGVLab47.6———
31LFM2.5-VL-3B+1 altLiquid AI47.2———
32LFM2-VL-3BLiquid AI46.4———
◆ 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: 32 models scored · 0 independently verified · 3 vendor cross-reference · 29 vendor-reported · 2 with source disagreement. How these tiers are assigned