AI Calculator Pro
Best deal

Best value LLMs: most intelligence per dollar

Every LLM value ranking divides quality by the price the lab publishes. But the same model is resold by dozens of providers, often for a fraction of that — so the list price answers “is this vendor’s price fair?”, not “where do I get the most intelligence per dollar?”. This table ranks 124 scored models by Arena score against the cheapest price a major provider charges for them — first-party vendors, the big clouds, and established hosts like OpenRouter, DeepInfra, Together, Fireworks and Groq. Drawn from 8,427 offerings across 2,007 models, rebuilt daily.

Best deal near the top

DeepSeek V4 Flash — 12,658 Elo per $/1M

The most Arena score per dollar among models within 5% of the best score we track. Arena 1424 at $0.1125 per 1M tokens blended — 57% below its $0.2625 list price, via DeepInfra.

Best deal at each quality level

The right answer depends on how smart you need it to be — score per dollar on its own always points at the cheapest weak model. These are the best deals inside each capability band.

Near the top
DeepSeek V4 Flash
Arena 1424 · $0.1125/1M via DeepInfra
within 5% of the best score we track (1497)
Strong enough for most work
Amazon Nova Lite
Arena 1298 · $0.1050/1M
within 15% of the best score
Budget
Amazon Nova Micro
Arena 1268 · $0.0613/1M
within 30% of the best score
124
Models ranked
28
Cheaper somewhere else
−89%
Biggest saving
11
On the frontier

4 models reach the price/quality frontier only when priced at what you can actually pay: DeepSeek V4 Flash, GLM-5.2, GLM-5.1, Qwen3.7 Max. On list prices alone they look like poor value.

1Amazon Nova MicroBest value
AWS Bedrock
1268$0.0613$0.0613—20,702
2Command R7B
Cohere
1262$0.0656$0.0656—19,230
3Phi 4
Microsoft
1217$0.0875$0.0875—13,909
4DeepSeek V4 FlashBest value
DeepSeek
1424$0.2625$0.1125
at DeepInfra
−57%12,658
5Amazon Nova LiteBest value
AWS Bedrock
1298$0.1050$0.1050—12,362
6Qwen3 32B
Alibaba
1340$1.22$0.1300
at DeepInfra
−89%10,308
7GPT-5 Nano
OpenAI
1320$0.1375$0.1375—9,600
8Step 3.5 Flash
StepFun
1404$0.1500$0.1500—9,360
9GLM-4.7-Flash
Z.ai (Zhipu AI)
1354$0.1450$0.1450—9,338
10Llama 4 Scout
Meta
1332$0.1675$0.1500
at OpenRouter
−10%8,880
11GLM-4.7-Flash (Bedrock)
AWS Bedrock
1354$0.1525$0.1525—8,879
12MiMo-V2.5Best value
Xiaomi
1426$0.1750$0.1750—8,149
13MiMo-V2-Omni
Xiaomi
1422$0.1750$0.1750—8,126
14MiMo-V2-Flash
Xiaomi
1395$0.1750$0.1750—7,971
15Gemini 2.0 Flash
Google
1354$0.1750$0.1750—7,737
16Gemini 2.5 Flash-Lite
Google
1330$0.1750$0.1750—7,600
17GPT-4.1 Nano
OpenAI
1300$0.1750$0.1750—7,429
18Llama-3.3-70B-Instruct
Meta
1275$0.1975$0.1975—6,456
19GLM-4.6V
Z.ai (Zhipu AI)
1375$0.4500$0.2238
at Zhipu AI
−50%6,145
20Mistral Small 4
Mistral AI
1325$0.2625$0.2625—5,048
21GPT-4o mini
OpenAI
1310$0.2625$0.2625—4,990
22DeepSeek V3.2
DeepSeek
1420$0.3150$0.2900
at DeepInfra
−8%4,897
23GLM-4.5-Air
Z.ai (Zhipu AI)
1383$0.4250$0.3100
at Novita AI
−27%4,461
24Command R
Cohere
1163$0.2625$0.2625—4,430
25Qwen3-Next 80B-A3B Instruct
Alibaba
1419$0.8750$0.3425
at DeepInfra
−61%4,143
26Qwen3.5 35B-A3B
Alibaba
1396$0.6875$0.3550
at DeepInfra
−48%3,932
27MiniMax-M2.7
MiniMax
1405$0.5250$0.3675
at OpenRouter
−30%3,823
28Llama 4 Maverick
Meta
1360$0.3775$0.3775—3,603
29Qwen Plus
Alibaba
1327$0.6000$0.3900
at OpenRouter
−35%3,403
30MiniMax-M2.5
MiniMax
1359$0.5250$0.4000
at DeepInfra
−24%3,398
31GLM-4.5V
Z.ai (Zhipu AI)
1334$0.9000$0.4483
at Zhipu AI
−50%2,976
32GPT-5.4 nano
OpenAI
1374$0.4625$0.4625—2,971
33MiniMax-M3Best value
MiniMax
1437$0.5250$0.4850
at DeepInfra
−8%2,963
34MiMo-V2.5-ProBest value
Xiaomi
1462$0.5437$0.5437—2,689
35MiniMax-M2.1
MiniMax
1391$0.5250$0.5250—2,650
36MiniMax M2.1 (Bedrock)
AWS Bedrock
1391$0.5250$0.5250—2,650
37MiMo-V2-Pro
Xiaomi
1436$0.5437$0.5437—2,641
38Qwen3.7 Plus
Alibaba
1459$0.7000$0.5600
at OpenRouter
−20%2,605
39MiniMax M2.5 (Bedrock)
AWS Bedrock
1359$0.5250$0.5250—2,589
40MiniMax-M2
MiniMax
1342$0.5250$0.5250—2,556
41MiniMax M2 (Bedrock)
AWS Bedrock
1342$0.5250$0.5250—2,556
42Devstral 2
Mistral AI
1340$0.5250$0.5250—2,552
43Gemini 3.1 Flash Lite
Google
1414$0.5625$0.5625—2,514
44Llama 3.3 70B
Meta
1315$0.6000$0.6000—2,192
45GPT-5 Mini
OpenAI
1374$0.6875$0.6875—1,999
46Qwen3.5 122B-A10B
Alibaba
1418$1.10$0.7150
at OpenRouter
−35%1,983
47Qwen3.6 Plus
Alibaba
1436$1.13$0.7312
at OpenRouter
−35%1,964
48GLM-4.7
Z.ai (Zhipu AI)
1436$1.00$0.7375
at DeepInfra
−26%1,947
49GPT-4.1 Mini
OpenAI
1345$0.7000$0.7000—1,921
50Mistral Large 3
Mistral AI
1431$0.7500$0.7500—1,908
51Qwen3.5 27B
Alibaba
1408$0.8250$0.8250—1,707
52Mistral Medium 3.1
Mistral AI
1350$0.8000$0.8000—1,688
53Gemini 2.5 Flash
Google
1417$0.8500$0.8500—1,667
54GLM-4.6
Z.ai (Zhipu AI)
1440$1.00$0.8750
at DeepInfra
−13%1,646
55Kimi K2.5
Moonshot AI
1445$1.20$0.9000
at DeepInfra
−25%1,606
56GLM-5
Z.ai (Zhipu AI)
1445$1.55$0.9700
at DeepInfra
−37%1,490
57DeepSeek V4 Pro
DeepSeek
1446$0.9900$0.9900—1,461
58GPT-3.5-turbo
OpenAI
1094$0.7500$0.7500—1,459
59GLM-4.7 (Bedrock)
AWS Bedrock
1436$1.00$1.00—1,436
60GLM-4.5
Z.ai (Zhipu AI)
1429$1.00$1.00—1,429
61Qwen3.5 397B-A17B
Alibaba
1438$1.35$1.09
at DeepInfra
−19%1,322
62Qwen/Qwen3-VL-235B-A22B-Instruct (Bedrock)
AWS Bedrock
1401$1.06$1.06—1,319
63Kimi K2 Thinking
Moonshot AI
1371$1.08$1.08—1,275
64Kimi K2 Thinking (Bedrock)
AWS Bedrock
1371$1.08$1.08—1,275
65GLM-5.2Best value
Z.ai (Zhipu AI)
1463$2.15$1.16
at DeepInfra
−46%1,258
66Kimi K2.5 (Bedrock)
AWS Bedrock
1445$1.20$1.20—1,204
67Qwen3-VL 235B-A22B
Alibaba
1401$1.22$1.22—1,144
68Qwen3 235B-A22B
Alibaba
1366$1.22$1.22—1,115
69Kimi K2.6
Moonshot AI
1455$1.71$1.34
at OpenRouter
−22%1,086
70GLM-5V-Turbo
Z.ai (Zhipu AI)
1442$1.90$1.38
at Zhipu AI
−27%1,043
71Amazon Nova Pro
AWS Bedrock
1348$1.40$1.40—963
72GLM-5.1Best value
Z.ai (Zhipu AI)
1466$2.15$1.57
at Zhipu AI
−27%935
73GLM-5 (Bedrock)
AWS Bedrock
1445$1.55$1.55—932
74Grok 4.3
xAI
1400$1.56$1.56—896
75GPT-5.4 mini
OpenAI
1413$1.69$1.69—837
76Qwen3.7 MaxBest value
Alibaba
1476$3.75$1.88
at Novita AI
−50%787
77Muse Spark 1.1Best value
Meta
1481$2.00$2.00—741
78Qwen3-Next 80B-A3B (Thinking)
Alibaba
1368$1.88$1.88—730
79o4-mini
OpenAI
1382$1.93$1.93—718
80Claude Haiku 4.5
Anthropic
1393$2.00$2.00—697
81o3-mini
OpenAI
1319$1.93$1.93—685
82Qwen3 Max
Alibaba
1439$2.40$2.40—600
83Qwen2.5 72B Instruct
Alibaba
1269$2.45$2.45—518
84Kimi K2 Thinking Turbo
Moonshot AI
1414$2.86$2.86—494
85Grok 4.5
xAI
1452$3.00$3.00—484
86Qwen Max
Alibaba
1282$2.80$2.80—458
87Qwen3-Coder 480B-A35B Instruct
Alibaba
1356$3.00$3.00—452
88Gemini 3.5 Flash
Google
1475$3.38$3.38—437
89Gemini 2.5 Pro
Google
1457$3.44$3.44—424
90GPT-5.1
OpenAI
1422$3.44$3.44—414
91GPT-5
OpenAI
1405$3.44$3.44—409
92o3
OpenAI
1422$3.50$3.50—406
93GPT-4.1
OpenAI
1372$3.50$3.50—392
94Claude Sonnet 5
Anthropic
1442$4.00$4.00—361
95Claude Sonnet 5 (Bedrock)
AWS Bedrock
1442$4.00$4.00—361
96Kimi K3
Moonshot AI
1473$6.00$4.15
at OpenRouter
−31%355
97Claude Sonnet 5 (India) (Bedrock)
AWS Bedrock
1442$4.40$4.40—328
98Command A
Cohere
1365$4.38$4.38—312
99GPT-4o
OpenAI
1355$4.38$4.38—310
100GPT-5.2
OpenAI
1411$4.81$4.81—293
101Amazon Nova Premier
AWS Bedrock
1400$5.00$5.00—280
102Command R+
Cohere
1204$4.38$4.38—275
103GPT-5.4
OpenAI
1453$5.63$5.63—258
104Kimi K3 (Bedrock)
AWS Bedrock
1473$6.00$6.00—246
105Claude Sonnet 4.6
Anthropic
1457$6.00$6.00—243
106Claude Sonnet 4.5 (latest)
Anthropic
1438$6.00$6.00—240
107GPT-5.4 (Bedrock)
AWS Bedrock
1453$6.19$6.19—235
108Anthropic Claude Sonnet 4.6 (Bedrock)
AWS Bedrock
1457$6.60$6.60—221
109GPT-5.6 Sol
OpenAI
1457$8.00$8.00—182
110GPT-5.6 Sol (Bedrock)
AWS Bedrock
1457$11.25$8.80
at Amazon Bedrock
−22%166
111Step 2 (16K)
StepFun
1321$8.02$8.02—165
112Claude Opus 4.6Best value
Anthropic
1497$10.00$10.00—150
113Claude Opus 4.7
Anthropic
1482$10.00$10.00—148
114Claude Opus 4.7 (Bedrock)
AWS Bedrock
1482$10.00$10.00—148
115Claude Opus 4.8
Anthropic
1454$10.00$10.00—145
116Claude Opus 4.8 (Bedrock)
AWS Bedrock
1454$10.00$10.00—145
117Claude Opus 4.5 (latest)
Anthropic
1450$10.00$10.00—145
118GPT-5.5
OpenAI
1464$11.25$11.25—130
119GPT-5.5 (Bedrock)
AWS Bedrock
1464$12.38$12.38—118
120Claude Fable 5
Anthropic
1493$20.00$20.00—75
121Claude Fable 5 (Bedrock)
AWS Bedrock
1493$20.00$20.00—75
122o1
OpenAI
1353$26.25$26.25—52
123Claude Opus 4.1 (latest)
Anthropic
1417$30.00$30.00—47
124GPT-4
OpenAI
1186$37.50$37.50—32

Best prices are third-party listings via models.dev, restricted to first-party vendors, the major clouds and established inference hosts — not rates we verify with each provider the way we verify first-party pricing. Cheaper obscure resellers exist in the data and are shown on each model page, but we will not rank on them. Use this as a shortlist to check: confirm quantisation, context limit, rate limits and data policy before switching, because a cheaper host is not always serving the same thing. Blended price weights input and output 3:1.

Intelligence scores are community Arena ratings from LMArena / arena.ai, used under CC BY 4.0. Snapshot last refreshed 18 November 2026. Scores are a relative signal, not an absolute measure of capability. A Measured score was voted on directly; an Estimated score is inherited from an identical base model (a regional/creator-prefixed hosting duplicate). See our methodology.

FAQ

What does “best value” mean here?

Arena score divided by price per 1M tokens, blended 3:1 input to output — but priced at the cheapest major provider selling that model, not at the model vendor's list price. A model is also flagged “Best value” when it sits on the price/quality frontier: nothing we track is both smarter and cheaper.

Why is this different from other value rankings?

Everyone else divides quality by the lab's list price. That answers whether a vendor's own price is fair, not what the most intelligence per dollar is. The same open-weight model is often resold 10–25× cheaper, so ranking on list price systematically undervalues models with a competitive hosting market.

Are these cheaper prices trustworthy?

They are third-party listings collected from models.dev, not rates we verify with each host the way we verify first-party pricing. We only quote first-party vendors, the major clouds and established inference hosts — the feed contains cheaper obscure resellers, and we deliberately do not rank on those, though you can see them on each model page. Treat the table as a shortlist to check: confirm quantisation, context limit, rate limits and data policy before switching, because a cheaper host is not always serving the same thing.

How often is it updated?

The site rebuilds daily against live pricing, so both the list prices and the third-party listings refresh every day, and the ranking moves with them.

Related

Estimates for planning only. Verify current pricing with each provider. Data rebuilt daily · https://aicalculatorpro.com/best-value-llm/