The AI Value Leaderboard: Intelligence per Dollar

Benchmarks tell you which model is smartest. Prices tell you which is cheapest. This table answers the question that matters: which model gives you the most capability for your money?

Value rankModelLabValue ScoreArena ScoreArena rankBlended $/1M
🥇gpt-oss-120bOpenAI2758.31366 ±4#68$0.06
🥈Qwen3 30B A3B Instruct 2507openQwen2175.91384 ±5#58$0.08
🥉DeepSeek V4 FlashopenDeepSeek2056.91431 ±5#31$0.11
4Gemma 4 26B A4B Google1839.21435 ±8#30$0.13
5Gemma 3 12BGoogle1789.31334 ±9#80$0.08
6Qwen3.5-FlashopenQwen1746.81399 ±4#53$0.11
7Gemma 3 27BGoogle15821358 ±4#72$0.1
8gpt-oss-20bOpenAI1545.41288 ±6#89$0.06
9Gemma 3 4BGoogle1452.81291 ±9#88$0.06
10Gemma 3n 4BGoogle14161306 ±5#86$0.08
11Gemma 4 31BGoogle13601441 ±8#23$0.18
12Qwen3 32BopenQwen1077.71340 ±9#78$0.13
13Gemini 2.5 Flash Lite Preview 09-2025Google1023.41379 ±3#60$0.18
14Gemini 2.5 Flash LiteGoogle962.91369 ±4#65$0.18
15GPT-5 NanoOpenAI874.91320 ±7#83$0.14
16DeepSeek V3.2openDeepSeek871.41424 ±4#33$0.26
17MiniMax M2.5openMiniMax757.11359 ±4#71$0.21
18MiniMax M2.7openMiniMax649.51405 ±4#49$0.32
19Qwen3 Next 80B A3B InstructopenQwen638.51419 ±5#37$0.34
20Qwen3 Next 80B A3B ThinkingopenQwen626.21368 ±6#66$0.27
21Qwen3 VL 235B A22B InstructopenQwen598.11421 ±7#35$0.37
22Mistral Small 3openMistral584.31234 ±6#101$0.06
23Qwen3.5-35B-A3BopenQwen5511396 ±4#54$0.36
24Qwen3 30B A3BopenQwen544.21317 ±5#85$0.22
25Llama 4 ScoutopenMeta538.71281 ±5#94$0.15
26DeepSeek V3.1 TerminusopenDeepSeek494.31418 ±10#39$0.44
27GPT-4.1 NanoOpenAI484.61285 ±8#92$0.18
28Llama 3.3 70B InstructopenMeta483.91275 ±3#96$0.16
29Qwen3.7 PlusopenQwen467.11462 ±6#10$0.56
30DeepSeek V4 ProopenDeepSeek459.61450 ±5#17$0.54
31MiniMax M3openMiniMax454.31439 ±6#26$0.52
32Qwen3 235B A22B Thinking 2507openQwen439.61414 ±7#44$0.49
33Qwen3.5-27BopenQwen3891409 ±4#47$0.54
34Gemini 3.1 Flash LiteGoogle382.91415 ±4#42$0.56
35GPT-5.4 NanoOpenAI3761374 ±4#61$0.46
36MiniMax M2.1openMiniMax363.61391 ±5#56$0.52
37Llama 4 MaverickopenMeta333.71288 ±4#90$0.26
38GPT-4o-mini (2024-07-18)OpenAI329.51287 ±4#91$0.26
39Qwen-PlusopenQwen324.41327 ±8#82$0.39
40Qwen3.6 PlusopenQwen323.61437 ±5#29$0.73
41MiniMax M2openMiniMax318.41342 ±8#76$0.45
42Kimi K2.5openMoonshot AI3111445 ±4#22$0.79
43Qwen3.5-122B-A10BopenQwen304.31418 ±4#38$0.72
44Qwen3.5 397B A17BopenQwen265.51439 ±4#24$0.9
45Gemini 2.5 FlashGoogle255.61417 ±2#41$0.85
46Qwen3 Coder 480B A35BopenQwen254.11356 ±5#73$0.62
47GPT-5 MiniOpenAI252.81374 ±5#62$0.69
48R1 0528openDeepSeek249.91428 ±6#32$0.91
49Qwen3 VL 235B A22B ThinkingopenQwen237.81401 ±7#52$0.85
50Gemini 3 Flash PreviewGoogle236.61466 ±4#8$1.13
51Qwen3 235B A22BopenQwen208.11366 ±5#67$0.8
52GPT-4.1 MiniOpenAI200.71341 ±4#77$0.7
53Kimi K2 ThinkingopenMoonshot AI199.21414 ±3#43$1.08
54Mistral Small 3.1 24BopenMistral193.81278 ±5#95$0.4
55Phi 4Microsoft1921217 ±5#104$0.09
56Nova 2 LiteAmazon191.41363 ±6#70$0.85
57Kimi K2.6openMoonshot AI189.21455 ±5#13$1.35
58Kimi K2 0711openMoonshot AI170.31371 ±5#64$1.00
59MiniMax M1openMiniMax167.41342 ±4#75$0.85
60Kimi K2 0905openMoonshot AI167.11380 ±6#59$1.08
61Grok 4.20xAI162.51454 ±4#15$1.56
62Grok 4.20 Multi-AgentxAI159.71450 ±4#19$1.56
63Qwen3 MaxopenQwen153.21439 ±5#25$1.56
64Llama 3.1 70B InstructopenMeta152.51261 ±4#100$0.4
65R1openDeepSeek150.21373 ±5#63$1.15
66Qwen3.7 MaxopenQwen146.71475 ±10#7$1.88
67Grok 4.3xAI129.41402 ±5#51$1.56
68GPT-5.4 MiniOpenAI126.31413 ±4#45$1.69
69Qwen3.6 Max PreviewopenQwen105.31447 ±8#20$2.34
70Claude Haiku 4.5Anthropic96.31393 ±3#55$2.00
71Gemini 3.5 FlashGoogle83.71482 ±6#3$3.38
72o4 MiniOpenAI79.31353 ±4#74$1.93
73Gemini 2.5 ProGoogle74.81457 ±2#11$3.44
74Mistral Medium 3.5openMistral73.21420 ±7#36$3.00
75GPT-5.1OpenAI64.61422 ±4#34$3.44
76Gemini 3.1 Pro Preview Custom ToolsGoogle62.11480 ±4#5$4.50
77Nano Banana Pro (Gemini 3 Pro Image)Google62.11479 ±4#6$4.50
78o3 MiniOpenAI61.91319 ±4#84$1.93
79Claude Sonnet 5Anthropic61.31445 ±9#21$4.00
80GPT-5OpenAI59.71405 ±4#48$3.44
81GPT-5 ChatOpenAI59.21404 ±4#50$3.44
82GPT-5.2 ChatOpenAI49.51438 ±4#27$4.81
83Gemma 2 27BGoogle48.51232 ±3#102$0.65
84GPT-5.4OpenAI451453 ±4#16$5.63
85GPT-5.2OpenAI43.91411 ±3#46$4.81
86Claude Sonnet 4.6Anthropic42.81457 ±4#12$6.00
87Claude Sonnet 4.5Anthropic39.71438 ±3#28$6.00
88GPT-5.3 ChatOpenAI39.11388 ±4#57$4.81
89Command ACohere301331 ±3#81$4.38
90Claude Opus 4.6Anthropic29.81498 ±4#1$10.00
91Claude Opus 4.7Anthropic28.21482 ±4#4$10.00
92Claude Opus 4.8Anthropic25.41454 ±6#14$10.00
93Claude Opus 4.5Anthropic251450 ±3#18$10.00
94GPT-5.5OpenAI23.41463 ±5#9$11.25
95Claude Sonnet 4Anthropic231338 ±4#79$6.00
96Mistral Large 2407openMistral22.11266 ±4#98$3.00
97GPT-4o (2024-08-06)OpenAI18.91283 ±4#93$4.38
98GPT-4.1OpenAI18.11264 ±4#99$3.50
99Claude Fable 5Anthropic14.81495 ±9#2$20.00
100GPT-4o (2024-05-13)OpenAI13.41301 ±3#87$7.50
101Claude Opus 4.1Anthropic7.21417 ±3#40$30.00
102Command R+ (08-2024)Cohere6.61229 ±7#103$4.38
103Claude Opus 4Anthropic5.51364 ±4#69$30.00
104GPT-4 TurboOpenAI4.81272 ±4#97$15.00

Methodology

Value Score = (Arena Score − 1200) ÷ blended price per 1M tokens, where blended price = 0.75 × input + 0.25 × output (a typical 3:1 token mix). The 1200-point baseline measures capability above a dated-model floor, so cheap-but-weak models can't win on price alone. A per-dollar metric still inherently favors efficient models over absolute frontier models — that's the point of this page; for raw capability, sort by Arena rank.

Capability data: Arena Scores © LMArena, licensed CC-BY-4.0, as of 2 Jul 2026. Prices: live market data, refreshed every 6 hours. No affiliate links on this page.