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
#1DeepSeek V4 Flash 0423openDeepSeek3484.71456 ±6#40$0.07
#2Qwen3 30B A3B Instruct 2507openQwen2578.61418 ±9#64$0.08
#3gpt-oss-120bOpenAI2565.11380 ±8#80$0.07
#4gpt-oss-20bOpenAI1969.11308 ±13#100$0.06
#5Qwen3.5-FlashopenQwen1875.21413 ±6#66$0.11
#6Gemma 4 26B A4B Google1776.71444 ±15#50$0.14
#7Gemma 4 31BGoogle1593.71455 ±15#41$0.16
#8Qwen3 32BopenQwen1216.21358 ±24#92$0.13
#9GPT-5 NanoOpenAI1106.21352 ±14#93$0.14
#10Gemma 3 12BGoogle10881282 ±23#107$0.08
#11Gemini 2.5 Flash LiteGoogle995.41374 ±8#84$0.18
#12Gemini 2.5 Flash Lite Preview 09-2025Google9801372 ±6#87$0.18
#13Gemma 3n 4BGoogle922.71269 ±10#109$0.08
#14DeepSeek V3.2openDeepSeek858.61449 ±6#47$0.29
#15Mistral Small 3openMistral803.51246 ±12#114$0.06
#16Gemma 3 27BGoogle710.11323 ±7#98$0.17
#17Qwen3 Next 80B A3B InstructopenQwen689.71441 ±9#51$0.35
#18Qwen3 30B A3BopenQwen6401338 ±9#94$0.22
#19GPT-4.1 NanoOpenAI606.31306 ±19#101$0.18
#20MiniMax M2.7openMiniMax601.91453 ±6#44$0.42
#21GPT-5.6 LunaOpenAI587.81465 ±9#34$0.45
#22Llama 4 ScoutopenMeta570.71286 ±9#105$0.15
#23MiniMax M3openMiniMax518.51472 ±7#29$0.52
#24Qwen3.7 PlusopenQwen498.61479 ±7#25$0.56
#25Gemma 3 4BGoogle491.21231 ±23#116$0.06
#26Qwen3 Next 80B A3B ThinkingopenQwen463.31391 ±11#77$0.41
#27DeepSeek V3.1 TerminusopenDeepSeek456.41407 ±21#70$0.45
#28Qwen3 Coder 480B A35BopenQwen447.41413 ±9#67$0.48
#29Llama 3.3 70B InstructopenMeta442.61269 ±7#111$0.16
#30GPT-5.4 NanoOpenAI441.51404 ±6#72$0.46
#31MiniMax M2.1openMiniMax424.61423 ±10#63$0.52
#32Qwen3.5-35B-A3BopenQwen419.41410 ±7#68$0.5
#33Qwen3.5-27BopenQwen419.21425 ±7#60$0.54
#34DeepSeek V4 Pro 0423openDeepSeek417.31471 ±6#30$0.65
#35DeepSeek V3.2 ExpopenDeepSeek411.51326 ±10#97$0.31
#36MiniMax M2openMiniMax389.71374 ±15#85$0.45
#37Qwen3 VL 235B A22B InstructopenQwen3791440 ±13#52$0.63
#38MiniMax M2.5openMiniMax377.61378 ±7#82$0.47
#39Qwen3.6 PlusopenQwen363.81466 ±6#32$0.73
#40Phi 4Microsoft362.31232 ±10#115$0.09
#41Gemini 3.1 Flash LiteGoogle354.11399 ±6#76$0.56
#42GPT-4o-mini (2024-07-18)OpenAI343.21290 ±7#104$0.26
#43Qwen3.5-122B-A10BopenQwen330.51436 ±7#55$0.72
#44Qwen-PlusopenQwen329.51329 ±18#96$0.39
#45Kimi K2.5openMoonshot AI304.11474 ±5#27$0.9
#46GPT-5 MiniOpenAI300.51407 ±8#69$0.69
#47Qwen3 235B A22B Thinking 2507openQwen300.11424 ±15#61$0.75
#48Gemini 3.5 Flash LiteGoogle299.51455 ±10#42$0.85
#49Llama 4 MaverickopenMeta292.31302 ±7#102$0.35
#50Mistral Small 3.1 24BopenMistral273.41310 ±8#99$0.4
#51Gemini 2.5 FlashGoogle262.91424 ±4#62$0.85
#52R1 0528openDeepSeek248.81427 ±11#59$0.91
#53GPT-4.1 MiniOpenAI239.31368 ±7#89$0.7
#54Kimi K2 ThinkingopenMoonshot AI235.61453 ±6#43$1.08
#55Qwen3 235B A22BopenQwen233.21386 ±9#79$0.8
#56Gemini 3 Flash PreviewGoogle233.21462 ±8#35$1.13
#57Nova 2 LiteAmazon220.11387 ±12#78$0.85
#58Qwen3.5 397B A17BopenQwen207.71465 ±5#33$1.27
#59Gemini 3.6 FlashGoogle196.91495 ±10#10$1.50
#60Kimi K2 0905openMoonshot AI185.81400 ±12#74$1.08
#61Kimi K2 0711openMoonshot AI177.61378 ±8#83$1.00
#62Qwen3 VL 235B A22B ThinkingopenQwen174.91427 ±14#58$1.30
#63Kimi K2.6openMoonshot AI167.41487 ±7#14$1.71
#64Grok 4.20xAI165.31458 ±6#37$1.56
#65MiniMax M1openMiniMax164.81359 ±8#91$0.96
#66Grok 4.20 Multi-AgentxAI164.51457 ±6#38$1.56
#67Qwen3 MaxopenQwen164.51457 ±8#39$1.56
#68Llama 3.1 70B InstructopenMeta150.21260 ±7#112$0.4
#69R1openDeepSeek149.41372 ±12#86$1.15
#70GPT-5.4 MiniOpenAI141.91440 ±6#53$1.69
#71Grok 4.3xAI138.11416 ±6#65$1.56
#72Qwen3.7 MaxopenQwen134.61498 ±18#9$2.21
#73Claude Haiku 4.5Anthropic125.51451 ±4#46$2.00
#74Qwen3.6 Max PreviewopenQwen116.11468 ±15#31$2.31
#75Qwen3.8 MaxopenQwen102.81508 ±13#6$3.00
#76Grok 4.5xAI93.81481 ±9#23$3.00
#77Grok 4.6xAI92.11476 ±27#26$3.00
#78o4 MiniOpenAI87.61369 ±7#88$1.93
#79Gemini 3.5 FlashGoogle87.21494 ±8#11$3.38
#80Mistral Medium 3.5openMistral86.81460 ±11#36$3.00
#81o3 MiniOpenAI84.51363 ±6#90$1.93
#82Mistral LargeopenMistral81.81445 ±5#49$3.00
#83Gemini 2.5 ProGoogle73.41452 ±4#45$3.44
#84GPT-5.6 SolOpenAI72.31489 ±9#12$4.00
#85Claude Sonnet 5Anthropic71.71487 ±8#13$4.00
#86GPT-5.1OpenAI68.81437 ±7#54$3.44
#87GPT-5OpenAI68.51435 ±8#56$3.44
#88GPT-5.6 TerraOpenAI63.11484 ±9#16$4.50
#89Gemini 3.1 Pro Preview Custom ToolsGoogle631484 ±5#17$4.50
#90Gemini 3.1 Pro PreviewGoogle62.91483 ±5#18$4.50
#91Nano Banana Pro (Gemini 3 Pro Image)Google62.81482 ±7#21$4.50
#92Nano Banana Pro (Gemini 3 Pro Image Preview)Google62.81483 ±7#19$4.50
#93GPT-5 ChatOpenAI581399 ±8#75$3.44
#94Kimi K3openMoonshot AI52.21513 ±11#5$6.00
#95GPT-5.2 ChatOpenAI51.21446 ±7#48$4.81
#96Claude Sonnet 4.6Anthropic50.71504 ±6#7$6.00
#97GPT-5.4OpenAI49.81480 ±6#24$5.63
#98GPT-5.2OpenAI48.21432 ±5#57$4.81
#99Claude Sonnet 4.5Anthropic47.51485 ±5#15$6.00
#100GPT-5.3 ChatOpenAI42.81406 ±7#71$4.81
#101Claude Opus 4.6Anthropic33.41534 ±5#1$10.00
#102Claude Opus 5Anthropic331530 ±9#2$10.00
#103Claude Opus 4.7Anthropic31.61516 ±6#4$10.00
#104Claude Opus 4.5Anthropic29.91499 ±5#8$10.00
#105Command ACohere29.91331 ±6#95$4.38
#106Claude Sonnet 4Anthropic29.91379 ±7#81$6.00
#107Claude Opus 4.8Anthropic28.21482 ±7#22$10.00
#108Mistral Large 2407openMistral25.71277 ±8#108$3.00
#109GPT-5.5OpenAI25.11483 ±6#20$11.25
#110GPT-4o (2024-08-06)OpenAI19.21284 ±8#106$4.38
#111Gemma 2 27BGoogle17.41211 ±6#117$0.65
#112Claude Fable 5Anthropic16.11522 ±9#3$20.00
#113GPT-4.1OpenAI15.51254 ±7#113$3.50
#114GPT-4o (2024-05-13)OpenAI131298 ±6#103$7.50
#115Claude Opus 4.1Anthropic9.11473 ±5#28$30.00
#116Claude Opus 4Anthropic6.71401 ±7#73$30.00
#117GPT-4 TurboOpenAI4.61269 ±7#110$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 12 Aug 2026. Prices: live market data, refreshed every 6 hours. No affiliate links on this page.