$

CheapTokenz

AI Token Price ComparisonCompare OpenAI, Grok, Claude, Gemini & More

Compare 66 AI models by token price, performance, and utility.
Find the cheapest AI API for your needs.

66
Total Models
15
Providers
$0.07
Cheapest: Llama 3.1 8B
95
Top MMLU: o3
76
Best Overall: Grok 4 Fast
View All ModelsCost Calculator

Model Price Comparison

Last updated: April 7, 2026 · Prices in USD per 1M tokens · Source: OpenRouter API

66 of 66 models
Grok 4 Fast
xAI · 2000K ctx
76
In
$0.20
Out
$0.50
MMLU
90
Speed + long context
Go →
Llama 3.1 405B
Meta · 128K ctx
72
In
$4.00
Out
$4.00
MMLU
87
Enterprise grade
Go →
GPT-5.4 Pro
OpenAI · 1050K ctx
72
In
$30.00
Out
$180.00
MMLU
94
Premium tasks
Go →
GPT-5.4
OpenAI · 1050K ctx
72
In
$2.50
Out
$15.00
MMLU
93
General purpose
Go →
GPT-4.1
OpenAI · 1048K ctx
71
In
$2.00
Out
$8.00
MMLU
89
Coding & analysis
Go →
Gemini 3.1 Pro
Google · 1049K ctx
71
In
$2.00
Out
$12.00
MMLU
91
Latest flagship
Go →
Llama 4 Maverick
Meta · 1049K ctx
71
In
$0.15
Out
$0.60
MMLU
88
Open source flagship
Go →
DeepSeek V3.2
DeepSeek · 164K ctx
69
In
$0.26
Out
$0.38
MMLU
88
Best value
Go →
Gemini 2.5 Pro
Google · 1049K ctx
68
In
$1.25
Out
$10.00
MMLU
88
Multimodal tasks
Go →
GPT-5.4 mini
OpenAI · 1050K ctx
68
In
$0.75
Out
$4.50
MMLU
87
Balanced mid-tier
Go →
Grok 4.20
xAI · 256K ctx
67
In
$2.00
Out
$6.00
MMLU
92
Multi-agent tasks
Go →
Nemotron 70B
NVIDIA · 131K ctx
67
In
$1.20
Out
$1.20
MMLU
83
Enterprise tuning
Go →
o3
OpenAI · 200K ctx
65
In
$2.00
Out
$8.00
MMLU
95
Deep reasoning
Go →
o4-mini
OpenAI · 200K ctx
65
In
$1.10
Out
$4.40
MMLU
94
Advanced reasoning
Go →
Grok 4
xAI · 256K ctx
65
In
$3.00
Out
$15.00
MMLU
93
Premium reasoning
Go →
GPT-5.2 Pro
OpenAI · 400K ctx
65
In
$21.00
Out
$168.00
MMLU
92.5
Pro reasoning
Go →
Claude Opus 4.6
Anthropic · 200K ctx
65
In
$5.00
Out
$25.00
MMLU
94.5
Flagship reasoning
Go →
Gemini 3 Flash
Google · 1049K ctx
65
In
$0.50
Out
$3.00
MMLU
84
Fast multimodal
Go →
Claude Opus 4.5
Anthropic · 200K ctx
65
In
$5.00
Out
$25.00
MMLU
94
Premium analysis
Go →
GPT-5.3
OpenAI · 400K ctx
65
In
$1.75
Out
$14.00
MMLU
91.5
Balanced performance
Go →
Claude Opus 4.1
Anthropic · 200K ctx
64
In
$15.00
Out
$75.00
MMLU
93.5
Deep research
Go →
o3-mini
OpenAI · 200K ctx
64
In
$1.10
Out
$4.40
MMLU
91
Reasoning tasks
Go →
GPT-5.2
OpenAI · 400K ctx
64
In
$1.75
Out
$14.00
MMLU
91
Advanced tasks
Go →
GPT-5.1
OpenAI · 400K ctx
64
In
$1.25
Out
$10.00
MMLU
90.5
Reliable workhorse
Go →
GPT-5
OpenAI · 400K ctx
64
In
$1.25
Out
$10.00
MMLU
90
Complex reasoning
Go →
Claude Sonnet 4.6
Anthropic · 200K ctx
64
In
$3.00
Out
$15.00
MMLU
91.5
Latest Sonnet
Go →
Claude Sonnet 4.5
Anthropic · 200K ctx
64
In
$3.00
Out
$15.00
MMLU
91
Coding & writing
Go →
DeepSeek R1-0528
DeepSeek · 164K ctx
63
In
$0.45
Out
$2.15
MMLU
91
Latest reasoning
Go →
Claude Sonnet 4
Anthropic · 200K ctx
63
In
$3.00
Out
$15.00
MMLU
90
Reliable all-rounder
Go →
GPT-4.1 mini
OpenAI · 1048K ctx
63
In
$0.40
Out
$1.60
MMLU
82
Balanced workloads
Go →
Claude 3.7 Sonnet
Anthropic · 200K ctx
63
In
$3.00
Out
$15.00
MMLU
89
Balanced performance
Go →
DeepSeek V3
DeepSeek · 164K ctx
62
In
$0.32
Out
$0.89
MMLU
86
Budget option
Go →
DeepSeek V3-0324
DeepSeek · 164K ctx
61
In
$0.20
Out
$0.77
MMLU
87
Proven stable
Go →
GPT-5.4 nano
OpenAI · 1050K ctx
61
In
$0.20
Out
$1.25
MMLU
82
Ultra low cost
Go →
Mistral Large 2512
Mistral · 128K ctx
61
In
$0.50
Out
$1.50
MMLU
86
Enterprise value
Go →
Grok 3
xAI · 131K ctx
61
In
$3.00
Out
$15.00
MMLU
88
Real-time knowledge
Go →
DeepSeek V3.1
DeepSeek · 164K ctx
60
In
$0.15
Out
$0.75
MMLU
86.5
Ultra cheap
Go →
Llama 4 Scout
Meta · 328K ctx
60
In
$0.08
Out
$0.30
MMLU
84
Efficient MoE
Go →
Qwen3.5 397B
Alibaba Cloud · 131K ctx
60
In
$0.39
Out
$2.34
MMLU
87.5
Qwen flagship
Go →
Mistral Large
Mistral · 128K ctx
60
In
$2.00
Out
$6.00
MMLU
85
European compliance
Go →
GPT-5 mini
OpenAI · 400K ctx
60
In
$0.25
Out
$2.00
MMLU
85
Cost-effective tasks
Go →
Qwen3 Max
Alibaba Cloud · 131K ctx
59
In
$0.78
Out
$3.90
MMLU
86.5
Complex reasoning
Go →
Llama 3.1 70B
Meta · 131K ctx
59
In
$0.40
Out
$0.40
MMLU
79
General use
Go →
Grok 3 Mini
xAI · 131K ctx
59
In
$0.30
Out
$0.50
MMLU
82
Budget reasoning
Go →
DeepSeek R1
DeepSeek · 64K ctx
59
In
$0.70
Out
$2.50
MMLU
90
Reasoning tasks
Go →
Qwen3 235B
Alibaba Cloud · 131K ctx
58
In
$0.45
Out
$1.82
MMLU
85
Complex reasoning
Go →
Grok Code Fast
xAI · 128K ctx
58
In
$0.20
Out
$1.50
MMLU
86
Code generation
Go →
Claude Haiku 4.5
Anthropic · 200K ctx
57
In
$1.00
Out
$5.00
MMLU
84
Fast & capable
Go →
ERNIE 4.5
Baidu · 128K ctx
57
In
$0.28
Out
$1.10
MMLU
84
Chinese NLP
Go →
Gemma 3 27B
Google · 131K ctx
57
In
$0.00
Out
$0.00
MMLU
74
Free open source
Go →
Gemini 2.5 Flash
Google · 1049K ctx
55
In
$0.30
Out
$2.50
MMLU
79
High volume
Go →
Llama 3.3 70B
Meta · 128K ctx
55
In
$0.10
Out
$0.32
MMLU
82
Best value Llama
Go →
Mistral Medium 3
Mistral · 128K ctx
55
In
$0.40
Out
$2.00
MMLU
83
Balanced mid-tier
Go →
GLM 4.5
Zhipu AI · 128K ctx
54
In
$0.60
Out
$2.20
MMLU
82
Enterprise apps
Go →
Qwen Max
Alibaba Cloud · 33K ctx
53
In
$1.04
Out
$4.16
MMLU
86
Complex tasks
Go →
MiniMax M1
MiniMax · 1000K ctx
52
In
$0.40
Out
$2.20
MMLU
77
Creative tasks
Go →
Gemini 2.0 Flash
Google · 1049K ctx
52
In
$0.10
Out
$0.40
MMLU
76
Cost efficiency
Go →
Kimi K2
Moonshot AI · 131K ctx
52
In
$0.55
Out
$2.20
MMLU
81
Long context
Go →
HunYuan Pro
Tencent · 128K ctx
50
In
$0.14
Out
$0.57
MMLU
80
Social integration
Go →
GPT-5 nano
OpenAI · 400K ctx
48
In
$0.05
Out
$0.40
MMLU
78
Cheapest GPT
Go →
Gemma 3 12B
Google · 33K ctx
47
In
$0.00
Out
$0.00
MMLU
70
Free lightweight
Go →
Qwen3 32B
Alibaba Cloud · 41K ctx
46
In
$0.08
Out
$0.24
MMLU
80
Multilingual tasks
Go →
Mixtral 8x22B
Mistral · 64K ctx
43
In
$2.00
Out
$6.00
MMLU
78
MoE architecture
Go →
Mistral Small 3
Mistral · 33K ctx
38
In
$0.05
Out
$0.08
MMLU
72
Ultra cheap
Go →
Phi-4
Microsoft · 16K ctx
38
In
$0.07
Out
$0.14
MMLU
75
Small model tasks
Go →
Llama 3.1 8B
Meta · 16K ctx
32
In
$0.02
Out
$0.05
MMLU
66
Simple tasks
Go →

🎯 How Utility Score Works

Utility Score = Weighted combination of 4 factors (0-100)

Performance
MMLU with Sigmoid curve — diminishing returns above 80
Price Efficiency
Lower cost = higher score (includes $0.01 base cost for free models)
Context Window
Log-scaled + step bonus: 128K (×1.1), 1M+ (×1.2)
Output Ratio
Input vs output price balance

💡 Select a profile above (Developer/Enterprise/Indie) or customize weights. Toggle "📊 Show Scores" to see per-factor breakdown.

⚠️ Price Disclaimer

Prices updated: April 7, 2026 from OpenRouter API. May vary by region.

Estimate Your Monthly Cost

How much would each model cost based on your actual usage?

100

~3000 messages/month

Gemini 2.5 Pro
$$23.25/mo
Gemini 2.0 Flash
$$1.14/mo
GPT-5.4 Pro
$$450.00/mo
GPT-5 mini
$$4.65/mo

💡 Estimate based on 70% input / 30% output ratio. Actual costs vary by use case.

About CheapTokenz

CheapTokenz is an AI token price comparison platform tracking 66 models across 15 providers including OpenAI, xAI Grok, Anthropic, Google, DeepSeek and Mistral. Find the cheapest AI API for your business.

Affiliate links help us keep the lights on — you pay no extra.

Frequently Asked Questions

What is the cheapest AI model in 2026?

The cheapest paid model is Llama 3.1 8B at $0.07/1M total. DeepSeek V3.2 at $0.64/1M offers the best performance-to-price ratio (88 MMLU). DeepSeek V3.1 at $0.90/1M (86.5 MMLU). Free: Gemma 3 27B (74 MMLU), Gemma 3 12B (70 MMLU). Note: Llama 3.3 70B is no longer free ($0.42/1M).

Which AI model has the highest MMLU score?

OpenAI o3 leads at 95 MMLU, followed by Claude Opus 4.6 (94.5), GPT-5.4 Pro and o4-mini (94). Budget top performer: DeepSeek R1-0528 (91 MMLU) at only $2.60/1M total.

How much does GPT-5.4 cost per token?

GPT-5.4: $2.50/1M input, $15.00/1M output ($17.50 total). GPT-5.4 Pro: $30/$180 ($210 total). GPT-5 mini: $0.25/$2.00 ($2.25 total). New: GPT-5 nano $0.05/$0.40 ($0.45) — cheapest GPT ever. GPT-5.4 mini: $0.75/$4.50 ($5.25).

Is DeepSeek cheaper than OpenAI?

Yes, 5-27x cheaper. DeepSeek V3.2: $0.26/$0.38 ($0.64 total) vs GPT-5.4: $2.50/$15 ($17.50 total). DeepSeek V3.1: $0.15/$0.75 ($0.90 total). DeepSeek R1: $0.70/$2.50 ($3.20 total) — 91 MMLU.

What is Grok 4 and how much does it cost?

Grok 4 (xAI): 93 MMLU, 256K context, $3/$15 ($18 total). Grok 4.20: $2/$6 ($8 total) for multi-agent tasks. Grok 4 Fast: 2M context, $0.20/$0.50 ($0.70 total). Grok Code Fast: $0.20/$1.50 for code gen.

How do Claude and GPT prices compare?

Claude Opus 4.6: $5/$25 ($30) — flagship tier. Claude Sonnet 4.6: $3/$15 ($18) ≈ GPT-5.4 ($17.50). Claude Haiku 4.5: $1/$5 ($6) vs GPT-5.4 nano: $0.20/$1.25 ($1.45). New Opus 4.5/4.6 are premium but capable.

What is the best free AI API?

Top free: Gemma 3 27B (Google, 74 MMLU, 131K ctx), Gemma 3 12B (Google, 70 MMLU, 32K ctx). ⚠️ Llama 3.3 70B no longer free on OpenRouter — now $0.10/$0.32 ($0.42/1M).

Which model has the longest context window?

Grok 4 Fast: 2M tokens. GPT-5.4 Pro/mini/nano, GPT-4.1, Gemini 2.5 Pro/Flash: ~1M tokens. MiniMax M1: 1M at $2.60/1M — cheapest long-context option.

Gemini 2.5 vs GPT-5.4: price comparison?

Gemini 2.5 Pro: $1.25/$10 ($11.25) — 36% cheaper than GPT-5.4 ($17.50). Gemini 3 Flash: $0.50/$3 ($3.50). Gemini 3.1 Pro: $2/$12 ($14) ≈ GPT-5.2.

What Chinese AI models are available?

DeepSeek V3.2 ($0.64, 88 MMLU), Qwen3.5 397B ($2.73, 87.5 MMLU), Qwen3 Max ($4.68, 86.5 MMLU), ERNIE 4.5 ($1.38, 84 MMLU), Kimi K2 ($2.75, 81 MMLU), HunYuan Pro ($0.71, 80 MMLU).