DarkRouter DarkRouter
Sign in
Catalogue · 68 models

Models and providers

Every model has several unofficial providers, and each has its own price for input, output, and cache. DarkRouter unifies them behind one API and always picks the cheapest working route. Expand any model to see all providers and their prices.

A

Anthropic

15
A

Claude Fable 5.1

context 1 M

Claude Fable 5.1 excels at agentic coding and long-running workflows, improving knowledge work and complex code refactoring.

INPUT / 1M
$1 $10
OUTPUT / 1M
$1.5 $50
model identifier
anthropic/claude-fable-5.1
Claude Hub · direct −96%
input$1
output$1.5
cache$0.2
Apixly · claude-zero-injection −45%
input$5.5
output$27.5
cache$0.138
PlusVibe · sub5max −36%
input$6.39
output$31.89
cache$0.645
Wellflow · claude −30%
input$7
output$35
cache$0.7
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Fable 5

context 1 M

Claude Fable 5 is a Mythos-class model for autonomous knowledge work and coding. It handles text, image, and file inputs to produce text outputs with reasoning support.

INPUT / 1M
$1 $10
OUTPUT / 1M
$1 $50
model identifier
anthropic/claude-fable-5
Claude Hub · direct −97%
input$1
output$1
cache$0.068
Apixly · claude-zero-injection −45%
input$5.5
output$27.5
cache$0.55
Wellflow · claude −30%
input$7
output$35
cache$0.7
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Opus 5.5

context 1 M

This model excels at complex reasoning, coding, and long-horizon agentic tasks, particularly with large codebases.

INPUT / 1M
$1 $4
OUTPUT / 1M
$1 $20
model identifier
anthropic/claude-opus-5.5
TSK · direct −92%
input$1
output$1
cache—
Wcnbai · ClaudeCode −80%
input$0.788
output$3.94
cache$0.039
Wellflow · claude −80%
input$0.8
output$4
cache$0.08
PlusVibe · sub5max −67%
input$1.3
output$6.54
cache$0.06
Apixly · claude code −65%
input$1.4
output$7
cache$0.07
Apixly · claude −60%
input$1.6
output$8
cache$0.08
Apixly · claude hybrid channel −51%
input$1.96
output$9.8
cache$0.098
Apixly · claude-zero-injection −45%
input$2.2
output$11
cache$0.11
OpenRouter Crystal · direct −1%
input$3.96
output$19.8
cache$0.198
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Sonnet 5.5

context 1 M

Claude Sonnet 5.5 is an upgraded model for everyday tasks, excelling at feature building and bug fixing.

INPUT / 1M
$0.5 $2
OUTPUT / 1M
$1 $10
model identifier
anthropic/claude-sonnet-5.5
TSK · direct −88%
input$0.5
output$1
cache$0.2
Wcnbai · ClaudeCode −80%
input$0.394
output$1.97
cache$0.02
Wellflow · claude −80%
input$0.4
output$2
cache$0.04
Wcnbai · ClaudeCode-Max −74%
input$0.525
output$2.63
cache$0.026
Wcnbai · ClaudeCode-Max（Pro） −74%
input$0.525
output$2.63
cache$0.026
PlusVibe · sub5max −67%
input$0.657
output$3.27
cache$0.036
Apixly · claude code −65%
input$0.7
output$3.5
cache$0.035
PlusVibe · direct −63%
input$0.74
output$3.72
cache$0.072
Apixly · claude −60%
input$0.8
output$4
cache$0.04
Apixly · claude hybrid channel −51%
input$0.98
output$4.9
cache$0.049
OpenRouter Crystal · direct −1%
input$1.98
output$9.9
cache$0.099
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Opus 5

context 1 M

Claude Opus 5 is a flagship model excelling at demanding reasoning, coding, and long-horizon agentic tasks, including software development and visual analysis.

INPUT / 1M
$0.5 $5
OUTPUT / 1M
$0.5 $25
model identifier
anthropic/claude-opus-5
Claude Hub · direct −97%
input$0.5
output$0.5
cache$0.5
Wellflow · claude −84%
input$0.8
output$4
cache$0.08
Wcnbai · ClaudeCode −80%
input$0.985
output$4.92
cache$0.098
Wcnbai · ClaudeCode-Max −74%
input$1.31
output$6.57
cache$0.131
Wcnbai · ClaudeCode-Max（Pro） −74%
input$1.31
output$6.57
cache$0.131
PlusVibe · direct −71%
input$1.47
output$7.34
cache$0.143
Apixly · claude code −65%
input$1.75
output$8.75
cache$0.175
PlusVibe · sub5max −64%
input$1.8
output$8.95
cache$0.179
Apixly · claude −60%
input$2
output$10
cache$0.2
Apixly · claude hybrid channel −51%
input$2.45
output$12.25
cache$0.245
Apixly · claude-zero-injection −45%
input$2.75
output$13.75
cache$0.275
Wcnbai · ccmax蒸馏 −43%
input$2.87
output$14.36
cache$0.287
Wcnbai · 正向claude −31%
input$3.45
output$17.24
cache$0.345
Wcnbai · Offical-Claude −14%
input$4.31
output$21.54
cache$0.431
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Sonnet 5

context 1 M

Sonnet 5 is Anthropic's most capable Sonnet model, excelling in coding, agents, and professional tasks with selectable reasoning effort levels.

INPUT / 1M
$0.4 $2
OUTPUT / 1M
$0.4 $10
model identifier
anthropic/claude-sonnet-5
Wellflow · claude −93%
input$0.4
output$0.4
cache$0.4
Claude Hub · direct −93%
input$0.4
output$0.4
cache$0.4
Wcnbai · ClaudeCode −80%
input$0.394
output$1.97
cache$0.039
Wcnbai · ClaudeCode-Max（Pro） −74%
input$0.525
output$2.63
cache$0.053
Apixly · claude code −65%
input$0.7
output$3.5
cache$0.07
Apixly · claude −60%
input$0.8
output$4
cache$0.08
Apixly · claude hybrid channel −51%
input$0.98
output$4.9
cache$0.098
PlusVibe · sub5max −48%
input$1.04
output$5.16
cache$0.107
Apixly · claude-zero-injection −45%
input$1.1
output$5.5
cache$0.11
Wcnbai · claude官混 −31%
input$1.38
output$6.89
cache$0.138
Wcnbai · 正向claude −31%
input$1.38
output$6.89
cache$0.138
Wcnbai · Offical-Claude −14%
input$1.72
output$8.62
cache$0.172
Wcnbai · AWS_Opus5_zl −11%
input$1.79
output$8.93
cache$0.179
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Haiku 4.5

context 200 K

Fastest, most affordable Claude for everyday tasks.

INPUT / 1M
$0.3 $1
OUTPUT / 1M
$0.3 $5
model identifier
anthropic/claude-haiku-4.5
Wellflow · claude −90%
input$0.3
output$0.3
cache$0.3
Wcnbai · ClaudeCode −80%
input$0.197
output$0.985
cache$0.02
Wcnbai · ClaudeCode-Max（Pro） −74%
input$0.263
output$1.31
cache$0.026
Wcnbai · ClaudeCode-Max（Pro） −74%
input$0.263
output$1.31
cache—
Wcnbai · ClaudeCode-Max −74%
input$0.263
output$1.31
cache$0.026
Apixly · claude code −65%
input$0.35
output$1.75
cache$0.035
Apixly · claude −60%
input$0.4
output$2
cache$0.04
Apixly · claude hybrid channel −51%
input$0.49
output$2.45
cache$0.049
Apixly · claude-zero-injection −45%
input$0.55
output$2.75
cache$0.055
Wcnbai · AWS-专供分组 −14%
input$0.862
output$4.31
cache$0.086
Wcnbai · Offical-Claude −14%
input$0.862
output$4.31
cache$0.086
Wcnbai · AWSB opus5 +15%
input$1.15
output$5.75
cache$0.115
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Sonnet 4.6

context 200 K

Balanced Claude with strong quality at a moderate price.

INPUT / 1M
$0.4 $3
OUTPUT / 1M
$0.4 $15
model identifier
anthropic/claude-sonnet-4.6
Wellflow · claude −96%
input$0.4
output$0.4
cache$0.4
Claude Hub · direct −96%
input$0.4
output$0.4
cache$0.4
Wcnbai · ClaudeCode −87%
input$0.394
output$1.97
cache$0.039
Wcnbai · ClaudeCode-Max（Pro） −82%
input$0.525
output$2.63
cache$0.053
Apixly · claude code −77%
input$0.7
output$3.5
cache$0.07
Apixly · claude −73%
input$0.8
output$4
cache$0.08
Apixly · claude hybrid channel −67%
input$0.98
output$4.9
cache$0.098
PlusVibe · sub5max −66%
input$1.04
output$5.16
cache$0.107
Apixly · claude-zero-injection −63%
input$1.1
output$5.5
cache$0.11
Wcnbai · claude官混 −54%
input$1.38
output$6.89
cache$0.138
Wcnbai · 正向claude −54%
input$1.38
output$6.89
cache$0.138
Wcnbai · Offical-Claude −43%
input$1.72
output$8.62
cache$0.172
Wcnbai · AWS_Opus5_zl −40%
input$1.79
output$8.93
cache$0.179
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Opus 4.8

context 200 K

Newest Opus — top-tier coding, analysis, and long-horizon tasks.

INPUT / 1M
$0.5 $5
OUTPUT / 1M
$0.5 $25
model identifier
anthropic/claude-opus-4.8
Claude Hub · direct −97%
input$0.5
output$0.5
cache$0.5
Wellflow · claude −84%
input$0.8
output$4
cache$0.08
Wcnbai · ClaudeCode −80%
input$0.985
output$4.92
cache$0.098
Wcnbai · ClaudeCode-Max −74%
input$1.31
output$6.57
cache$0.131
Wcnbai · ClaudeCode-Max（Pro） −74%
input$1.31
output$6.57
cache$0.131
PlusVibe · direct −71%
input$1.47
output$7.34
cache$0.143
Apixly · claude code −65%
input$1.75
output$8.75
cache$0.175
PlusVibe · sub5max −64%
input$1.8
output$8.95
cache$0.179
Apixly · claude −60%
input$2
output$10
cache$0.2
Apixly · claude hybrid channel −51%
input$2.45
output$12.25
cache$0.245
Apixly · claude-zero-injection −45%
input$2.75
output$13.75
cache$0.275
Wcnbai · ccmax蒸馏 −43%
input$2.87
output$14.36
cache$0.287
Wcnbai · 正向claude −31%
input$3.45
output$17.24
cache$0.345
Wcnbai · Offical-Claude −14%
input$4.31
output$21.54
cache$0.431
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Opus 4.7

context 200 K

Most capable Claude for deep reasoning and agentic work.

INPUT / 1M
$0.5 $5
OUTPUT / 1M
$0.5 $25
model identifier
anthropic/claude-opus-4.7
Claude Hub · direct −97%
input$0.5
output$0.5
cache$0.5
Wellflow · claude −84%
input$0.8
output$4
cache$0.08
Wcnbai · ClaudeCode −80%
input$0.985
output$4.92
cache$0.098
Wcnbai · ClaudeCode-Max −74%
input$1.31
output$6.57
cache$0.131
Wcnbai · ClaudeCode-Max（Pro） −74%
input$1.31
output$6.57
cache$0.131
PlusVibe · direct −71%
input$1.47
output$7.34
cache$0.143
Apixly · claude code −65%
input$1.75
output$8.75
cache$0.175
PlusVibe · sub5max −64%
input$1.8
output$8.95
cache$0.179
Apixly · claude −60%
input$2
output$10
cache$0.2
Apixly · claude hybrid channel −51%
input$2.45
output$12.25
cache$0.245
Apixly · claude-zero-injection −45%
input$2.75
output$13.75
cache$0.275
Wcnbai · ccmax蒸馏 −43%
input$2.87
output$14.36
cache$0.287
Wcnbai · 正向claude −31%
input$3.45
output$17.24
cache$0.345
Wcnbai · Offical-Claude −14%
input$4.31
output$21.54
cache$0.431
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Opus 4.6

INPUT / 1M
$0.5 $5
OUTPUT / 1M
$0.5 $25
model identifier
anthropic/claude-opus-4.6
Claude Hub · direct −97%
input$0.5
output$0.5
cache$0.5
Wellflow · claude −84%
input$0.8
output$4
cache$0.08
Wcnbai · ClaudeCode −80%
input$0.985
output$4.92
cache$0.098
Wcnbai · ClaudeCode-Max −74%
input$1.31
output$6.57
cache$0.131
Wcnbai · ClaudeCode-Max（Pro） −74%
input$1.31
output$6.57
cache$0.131
PlusVibe · direct −71%
input$1.47
output$7.34
cache$0.143
Apixly · claude code −65%
input$1.75
output$8.75
cache$0.175
PlusVibe · sub5max −64%
input$1.8
output$8.95
cache$0.179
Apixly · claude −60%
input$2
output$10
cache$0.2
Apixly · claude hybrid channel −51%
input$2.45
output$12.25
cache$0.245
Apixly · claude-zero-injection −45%
input$2.75
output$13.75
cache$0.275
Wcnbai · ccmax蒸馏 −43%
input$2.87
output$14.36
cache$0.287
Wcnbai · 正向claude −31%
input$3.45
output$17.24
cache$0.345
Wcnbai · Offical-Claude −14%
input$4.31
output$21.54
cache$0.431
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Opus 4.5

INPUT / 1M
$0.5 $5
OUTPUT / 1M
$0.5 $25
model identifier
anthropic/claude-opus-4.5
Claude Hub · direct −97%
input$0.5
output$0.5
cache$0.5
Wellflow · claude −84%
input$0.8
output$4
cache$0.08
Wcnbai · ClaudeCode −80%
input$0.985
output$4.92
cache$0.098
Wcnbai · ClaudeCode-Max −74%
input$1.31
output$6.57
cache$0.131
Wcnbai · ClaudeCode-Max（Pro） −74%
input$1.31
output$6.57
cache$0.131
PlusVibe · direct −71%
input$1.47
output$7.34
cache$0.143
Apixly · claude code −65%
input$1.75
output$8.75
cache$0.175
PlusVibe · sub5max −64%
input$1.8
output$8.95
cache$0.179
Apixly · claude −60%
input$2
output$10
cache$0.2
Apixly · claude hybrid channel −51%
input$2.45
output$12.25
cache$0.245
Apixly · claude-zero-injection −45%
input$2.75
output$13.75
cache$0.275
Wcnbai · ccmax蒸馏 −43%
input$2.87
output$14.36
cache$0.287
Wcnbai · 正向claude −31%
input$3.45
output$17.24
cache$0.345
Wcnbai · Offical-Claude −14%
input$4.31
output$21.54
cache$0.431
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Sonnet 4.5

context 1 M

Claude Sonnet 4.5 is an advanced model optimized for agentic workflows and coding. It excels at complex coding tasks and benchmarks.

INPUT / 1M
$0.4 $3
OUTPUT / 1M
$0.4 $15
model identifier
anthropic/claude-sonnet-4.5
Wellflow · claude −96%
input$0.4
output$0.4
cache$0.4
Claude Hub · direct −96%
input$0.4
output$0.4
cache$0.4
Wcnbai · ClaudeCode −87%
input$0.394
output$1.97
cache$0.039
Wcnbai · ClaudeCode-Max（Pro） −82%
input$0.525
output$2.63
cache$0.053
Apixly · claude code −77%
input$0.7
output$3.5
cache$0.07
Apixly · claude −73%
input$0.8
output$4
cache$0.08
Apixly · claude hybrid channel −67%
input$0.98
output$4.9
cache$0.098
PlusVibe · sub5max −66%
input$1.04
output$5.16
cache$0.107
Apixly · claude-zero-injection −63%
input$1.1
output$5.5
cache$0.11
Wcnbai · claude官混 −54%
input$1.38
output$6.89
cache$0.138
Wcnbai · 正向claude −54%
input$1.38
output$6.89
cache$0.138
Wcnbai · Offical-Claude −43%
input$1.72
output$8.62
cache$0.172
Wcnbai · AWS_Opus5_zl −40%
input$1.79
output$8.93
cache$0.179
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

anthropic/claude-sonnet-4

INPUT / 1M
$0.4 $3
OUTPUT / 1M
$0.4 $15
model identifier
anthropic/claude-sonnet-4
Wellflow · claude −96%
input$0.4
output$0.4
cache$0.4
Claude Hub · direct −96%
input$0.4
output$0.4
cache$0.4
Wcnbai · ClaudeCode −87%
input$0.394
output$1.97
cache$0.039
Wcnbai · ClaudeCode-Max（Pro） −82%
input$0.525
output$2.63
cache$0.053
Apixly · claude code −77%
input$0.7
output$3.5
cache$0.07
Apixly · claude −73%
input$0.8
output$4
cache$0.08
Apixly · claude hybrid channel −67%
input$0.98
output$4.9
cache$0.098
PlusVibe · sub5max −66%
input$1.04
output$5.16
cache$0.107
Apixly · claude-zero-injection −63%
input$1.1
output$5.5
cache$0.11
Wcnbai · claude官混 −54%
input$1.38
output$6.89
cache$0.138
Wcnbai · 正向claude −54%
input$1.38
output$6.89
cache$0.138
Wcnbai · Offical-Claude −43%
input$1.72
output$8.62
cache$0.172
Wcnbai · AWS_Opus5_zl −40%
input$1.79
output$8.93
cache$0.179
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
A

Claude Haiku 5.5

context 1 M

Claude Haiku 5.5 is a fast, cost-effective model for high-volume tasks like summarization and subagents, with improved coding and computer use capabilities.

INPUT / 1M
$0.036 $0.1
OUTPUT / 1M
$0.155 $0.5
model identifier
anthropic/claude-haiku-5.5
PlusVibe · sub5max −68%
input$0.036
output$0.155
cache$0.004
Wcnbai · ClaudeCode +97%
input$0.197
output$0.985
cache$0.02
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

OpenAI

17
O

GPT-6 Astra

context 1.1 M

GPT-6 Astra is a flagship model for demanding end-to-end tasks, excelling in advanced analysis, software engineering, research, and document creation.

INPUT / 1M
$0.4 $10
OUTPUT / 1M
$0.4 $50
model identifier
openai/gpt-6-astra
Claude Hub · direct −99%
input$0.4
output$0.4
cache$0.4
Apixly · gpt −90%
input$1
output$5
cache$0.1
PlusVibe · direct −88%
input$1.23
output$6.15
cache$0.119
Apixly · Azure-gpt −45%
input$5.5
output$27.5
cache$0.55
OpenRouter Crystal · direct −1%
input$9.9
output$49.5
cache$0.99
Wcnbai · AZ +9%
input$13.79
output$51.71
cache$1.38
Wcnbai · az定制 +56%
input$19.7
output$73.87
cache$1.97
Wcnbai · AZ_USEast +100%
input$25.28
output$94.8
cache$2.53
Wcnbai · az gpt Coding +108%
input$26.26
output$98.49
cache$2.63
Wcnbai · AZ蒸馏 +160%
input$32.83
output$123.11
cache$3.28
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-6.1 Sol

context 1.1 M

A capable coding and professional assistant, GPT-6.1 Sol excels at agentic tasks and handling extensive documentation.

INPUT / 1M
$0.2 $2
OUTPUT / 1M
$0.2 $10
model identifier
openai/gpt-6.1-sol
Subscriptions · direct −97%
input$0.2
output$0.2
cache$0.02
PlusVibe · direct −91%
input$0.17
output$0.85
cache$0.012
PlusVibe · direct −91%
input$0.17
output$0.85
cache$0.008
Apixly · gpt −90%
input$0.2
output$1
cache$0.01
Wellflow · direct −87%
input$0.27
output$1.32
cache$0.01
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-6 Sol

context 1.1 M

GPT-6 Sol is a cost-efficient, high-end model for demanding professional tasks. It offers strong performance at a mid-tier price point.

INPUT / 1M
$0.2 $2
OUTPUT / 1M
$0.2 $10
model identifier
openai/gpt-6-sol
Subscriptions · direct −97%
input$0.2
output$0.2
cache—
Claude Hub · direct −95%
input$0.3
output$0.3
cache$0.3
Wellflow · direct −94%
input$0.13
output$0.63
cache$0.01
PlusVibe · direct −91%
input$0.17
output$0.85
cache$0.012
Apixly · gpt −90%
input$0.2
output$1
cache$0.02
Apixly · Azure-gpt −45%
input$1.1
output$5.5
cache$0.11
Wcnbai · az定制 +56%
input$3.94
output$14.77
cache$0.394
Wcnbai · az gpt Coding +108%
input$5.25
output$19.7
cache$0.525
Wcnbai · AZ蒸馏 +160%
input$6.57
output$24.62
cache$0.657
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-6 Luna

context 1.1 M

GPT-6 Luna is a fast, cost-efficient model ideal for high-volume, latency-sensitive tasks like chat and classification.

INPUT / 1M
$0.009 $0.1
OUTPUT / 1M
$0.043 $0.5
model identifier
openai/gpt-6-luna
PlusVibe · direct −91%
input$0.009
output$0.043
cache$0.001
Wellflow · direct −58%
input$0.04
output$0.21
cache$0.03
PlusVibe · direct −54%
input$0.045
output$0.228
cache$0.005
Claude Hub · direct −33%
input$0.2
output$0.2
cache$0.2
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-5.6 Sol

context 1.1 M

This model excels at complex reasoning, coding, and agentic workflows, particularly with command-line and multi-step coding tasks.

INPUT / 1M
$0.22 $4
OUTPUT / 1M
$1.07 $20
model identifier
openai/gpt-5.6-sol
Wellflow · direct −95%
input$0.22
output$1.07
cache$0.02
Apixly · gpt −90%
input$0.4
output$2
cache$0.04
Apixly · Azure-gpt −45%
input$2.2
output$11
cache$0.22
Wcnbai · AZ +9%
input$5.52
output$20.68
cache$0.552
Wcnbai · az定制 +56%
input$7.88
output$29.55
cache$0.788
Wcnbai · oai-image +98%
input$9.98
output$37.43
cache$0.998
Wcnbai · az gpt Coding +108%
input$10.51
output$39.4
cache$1.05
Wcnbai · oai +155%
input$12.87
output$48.26
cache$1.29
Wcnbai · AZ蒸馏 +160%
input$13.13
output$49.24
cache$1.31
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-5.6 Terra

context 1.1 M

A balanced model for everyday coding, reasoning, and agentic tasks.

INPUT / 1M
$0.13 $2
OUTPUT / 1M
$0.63 $12
model identifier
openai/gpt-5.6-terra
Wellflow · direct −95%
input$0.13
output$0.63
cache$0.01
Claude Hub · direct −94%
input$0.4
output$0.4
cache$0.4
Apixly · gpt −90%
input$0.2
output$1.2
cache$0.02
Apixly · Azure-gpt −45%
input$1.1
output$6.6
cache$0.11
Wcnbai · AZ +8%
input$2.76
output$12.41
cache$0.276
Wcnbai · az定制 +55%
input$3.94
output$17.73
cache$0.394
Wcnbai · oai-image +96%
input$4.99
output$22.46
cache$0.499
Wcnbai · az gpt Coding +106%
input$5.25
output$23.64
cache$0.525
Wcnbai · oai +153%
input$6.43
output$28.96
cache$0.643
Wcnbai · AZ蒸馏 +158%
input$6.57
output$29.55
cache$0.657
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-5.6 Luna

context 1.1 M

A fast, cost-efficient model for high-volume, latency-sensitive tasks like chat and classification. It offers capable reasoning for lightweight agentic workflows.

INPUT / 1M
$0.009 $0.2
OUTPUT / 1M
$0.043 $1.2
model identifier
openai/gpt-5.6-luna
PlusVibe · direct −96%
input$0.009
output$0.043
cache$0.001
Wellflow · direct −82%
input$0.04
output$0.21
cache$0.03
PlusVibe · direct −80%
input$0.045
output$0.228
cache$0.005
Claude Hub · direct −71%
input$0.2
output$0.2
cache$0.2
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-5.5

context 1 M

Latest frontier model with the strongest general capabilities.

INPUT / 1M
$0.3 $5
OUTPUT / 1M
$0.3 $30
model identifier
openai/gpt-5.5
Wellflow · direct −98%
input$0.3
output$0.3
cache$0.3
Apixly · Azure-gpt −45%
input$2.75
output$16.5
cache$0.275
Wcnbai · AZ +8%
input$6.89
output$31.02
cache$0.689
Wcnbai · az定制 +55%
input$9.85
output$44.32
cache$0.985
Wcnbai · az gpt Coding +106%
input$13.13
output$59.09
cache$1.31
Wcnbai · oai +153%
input$16.09
output$72.39
cache$1.61
Wcnbai · AZ蒸馏 +158%
input$16.41
output$73.87
cache$1.64
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-5.3 Codex

context 400 K

Tuned for coding agents — strong at long edits and tool calling.

INPUT / 1M
$0.3 $1.75
OUTPUT / 1M
$0.3 $14
model identifier
openai/gpt-5.3-codex
Wellflow · direct −96%
input$0.3
output$0.3
cache$0.3
Apixly · Azure-gpt +22%
input$2.75
output$16.5
cache$0.275
Wcnbai · AZ +141%
input$6.89
output$31.02
cache$0.689
Wcnbai · az定制 +244%
input$9.85
output$44.32
cache$0.985
Wcnbai · az gpt Coding +359%
input$13.13
output$59.09
cache$1.31
Wcnbai · oai +462%
input$16.09
output$72.39
cache$1.61
Wcnbai · AZ蒸馏 +473%
input$16.41
output$73.87
cache$1.64
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-5.4

context 1 M

Higher-quality reasoning for complex, multi-step tasks.

INPUT / 1M
$0.3 $2.5
OUTPUT / 1M
$0.3 $15
model identifier
openai/gpt-5.4
Wellflow · direct −97%
input$0.3
output$0.3
cache$0.3
Apixly · Azure-gpt −45%
input$1.38
output$8.25
cache$0.138
Wcnbai · AZ +8%
input$3.45
output$15.51
cache$0.345
Wcnbai · az定制 +55%
input$4.92
output$22.16
cache$0.492
Wcnbai · oai +153%
input$8.04
output$36.19
cache$0.804
Wcnbai · AZ蒸馏 +158%
input$8.21
output$36.93
cache$0.821
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-5.4 Nano

context 400 K

A fast, cost-efficient model for high-volume tasks, supporting text and image inputs with low latency.

INPUT / 1M
$0.069 $0.2
OUTPUT / 1M
$0.431 $1.25
model identifier
openai/gpt-5.4-nano
Wcnbai · AZ −66%
input$0.069
output$0.431
cache$0.007
Wcnbai · az定制 −51%
input$0.098
output$0.616
cache$0.01
Apixly · Azure-gpt −45%
input$0.11
output$0.688
cache$0.011
Wcnbai · oai −20%
input$0.161
output$1.01
cache$0.016
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-5.2

context 400 K

Balanced flagship for everyday reasoning, chat, and tool use.

INPUT / 1M
$0.603 $1.75
OUTPUT / 1M
$4.83 $14
model identifier
openai/gpt-5.2
Wcnbai · AZ −66%
input$0.603
output$4.83
cache—
Wcnbai · az定制 −51%
input$0.862
output$6.89
cache—
Wcnbai · oai −20%
input$1.41
output$11.26
cache—
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-5 Mini

context 400 K

GPT-5 Mini is a smaller, faster, and cheaper version of GPT-5, ideal for lighter reasoning tasks while retaining instruction-following and safety features.

INPUT / 1M
$0.086 $0.25
OUTPUT / 1M
$0.689 $2
model identifier
openai/gpt-5-mini
Wcnbai · AZ −66%
input$0.086
output$0.689
cache$0.009
Wcnbai · az定制 −51%
input$0.123
output$0.985
cache$0.012
Apixly · Azure-gpt −45%
input$0.138
output$1.1
cache$0.014
Wcnbai · oai −20%
input$0.201
output$1.61
cache$0.02
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-4.1

context 1 M

This advanced LLM excels at complex instruction following, software engineering tasks, and reasoning over very long contexts.

INPUT / 1M
$1.1 $2
OUTPUT / 1M
$4.4 $8
model identifier
openai/gpt-4.1
Apixly · Azure-gpt −45%
input$1.1
output$4.4
cache$0.275
Wcnbai · oai −20%
input$1.61
output$6.43
cache$0.402
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-4.1 Mini

context 1 M

A mid-sized model offering GPT-4o-level performance with lower latency and cost. It excels with a large context window and strong performance on challenging tasks.

INPUT / 1M
$0.138 $0.4
OUTPUT / 1M
$0.552 $1.6
model identifier
openai/gpt-4.1-mini
Wcnbai · AZ −66%
input$0.138
output$0.552
cache$0.034
Wcnbai · az定制 −51%
input$0.197
output$0.788
cache$0.049
Apixly · Azure-gpt −45%
input$0.22
output$0.88
cache$0.055
Wcnbai · oai −20%
input$0.322
output$1.29
cache$0.08
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-4o

context 130 K

This is OpenAI's latest multimodal model, accepting text and image inputs to generate text outputs. It offers GPT-4 Turbo intelligence with enhanced speed.

INPUT / 1M
$0.862 $2.5
OUTPUT / 1M
$3.45 $10
model identifier
openai/gpt-4o
Wcnbai · AZ −66%
input$0.862
output$3.45
cache$0.431
Wcnbai · az定制 −51%
input$1.23
output$4.92
cache$0.616
Apixly · Azure-gpt −45%
input$1.38
output$5.5
cache$0.688
Wcnbai · oai −20%
input$2.01
output$8.04
cache$1.01
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
O

GPT-4o-mini

context 130 K

This advanced small model handles text and image inputs, generating text outputs. It offers powerful capabilities at a significantly reduced cost.

INPUT / 1M
$0.052 $0.15
OUTPUT / 1M
$0.207 $0.6
model identifier
openai/gpt-4o-mini
Wcnbai · AZ −66%
input$0.052
output$0.207
cache$0.026
Wcnbai · az定制 −51%
input$0.074
output$0.295
cache$0.037
Apixly · Azure-gpt −45%
input$0.082
output$0.33
cache$0.041
Wcnbai · oai −20%
input$0.121
output$0.483
cache$0.06
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Google Gemini

13
G

Gemini 3.8 Flash

context 1 M

Gemini 3.8 Flash is Google's intelligent Flash model, excelling in software engineering, agentic tasks, and multi-step reasoning.

INPUT / 1M
$0.123 $0.825
OUTPUT / 1M
$0.616 $4.12
model identifier
google/gemini-3.8-flash
Wcnbai · Gemini CLI −85%
input$0.123
output$0.616
cache$0.012
Wcnbai · Vertex −73%
input$0.222
output$1.11
cache$0.022
Apixly · gemini cli −60%
input$0.33
output$1.65
cache$0.033
Wcnbai · Gemini优质文本 −55%
input$0.369
output$1.85
cache$0.037
Apixly · gemini vertex −50%
input$0.412
output$2.06
cache$0.041
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 3.7 Flash

context 1 M

A fast multimodal model for agentic workflows, coding, and complex multi-step reasoning. It excels at responsive performance and reliable multi-step tasks.

INPUT / 1M
$0.123 $0.75
OUTPUT / 1M
$0.616 $3.75
model identifier
google/gemini-3.7-flash
Wcnbai · Gemini CLI −84%
input$0.123
output$0.616
cache$0.012
Wcnbai · Vertex −70%
input$0.222
output$1.11
cache$0.022
Apixly · gemini cli −60%
input$0.3
output$1.5
cache$0.03
Wcnbai · Gemini优质文本 −51%
input$0.369
output$1.85
cache$0.037
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 3.6 Flash

context 1 M

Gemini 3.6 Flash is a fast, efficient model for coding and agentic workflows, producing polished outputs with minimal edits.

INPUT / 1M
$0.123 $0.75
OUTPUT / 1M
$0.616 $3.75
model identifier
google/gemini-3.6-flash
Wcnbai · Gemini CLI −84%
input$0.123
output$0.616
cache$0.012
Wcnbai · Vertex −70%
input$0.222
output$1.11
cache$0.022
Apixly · gemini cli −60%
input$0.3
output$1.5
cache$0.03
Wcnbai · Gemini优质文本 −51%
input$0.369
output$1.85
cache$0.037
Apixly · image
input$0.75
output$3.75
cache$0.075
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 3.5 Flash

context 1 M

Latest Flash generation, optimised for speed and scale.

INPUT / 1M
$0.15 $1.5
OUTPUT / 1M
$0.15 $9
model identifier
google/gemini-3.5-flash
Wellflow · direct −97%
input$0.15
output$0.15
cache$0.15
Wcnbai · Gemini CLI −84%
input$0.246
output$1.48
cache$0.025
Wcnbai · Vertex −70%
input$0.443
output$2.66
cache$0.044
Apixly · gemini cli −60%
input$0.6
output$3.6
cache$0.06
Wcnbai · Gemini优质文本 −51%
input$0.739
output$4.43
cache$0.074
Apixly · gemini vertex −50%
input$0.75
output$4.5
cache$0.075
Apixly · image
input$1.5
output$9
cache$0.15
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 3.1 Pro

context 2 M

Refined Gemini 3 Pro with an even larger 2M-token context.

INPUT / 1M
$0.15 $2
OUTPUT / 1M
$0.9 $12
model identifier
google/gemini-3.1-pro-preview
Wellflow · direct −93%
input$0.15
output$0.9
cache$0.03
Apixly · gemini cli −60%
input$0.8
output$4.8
cache$0.08
Apixly · gemini vertex −50%
input$1
output$6
cache$0.1
Wcnbai · Vertex −7%
input$2.36
output$10.64
cache$0.236
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 3.1 Flash Lite

context 1 M

Gemini 3.1 Flash Lite is a high-efficiency multimodal model for low-latency, high-volume tasks. It handles text, image, video, audio, and PDF inputs.

INPUT / 1M
$0.074 $0.25
OUTPUT / 1M
$0.443 $1.5
model identifier
google/gemini-3.1-flash-lite
Wcnbai · Vertex −70%
input$0.074
output$0.443
cache$0.007
PlusVibe · direct −6%
input$0.234
output$1.41
cache$0.012
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 3.1 Flash Lite Preview

context 1 M

A high-efficiency model for high-volume use cases, offering improved quality and performance.

INPUT / 1M
$0.041 $0.25
OUTPUT / 1M
$0.246 $1.5
model identifier
google/gemini-3.1-flash-lite-preview
Wcnbai · Gemini CLI −84%
input$0.041
output$0.246
cache$0.004
Wcnbai · Vertex −70%
input$0.074
output$0.443
cache$0.007
Wcnbai · Gemini优质文本 −51%
input$0.123
output$0.739
cache$0.012
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 3 Flash Preview

context 1 M

A fast, cost-effective model for agentic workflows, chat, and coding, offering near Pro-level reasoning.

INPUT / 1M
$0.082 $0.5
OUTPUT / 1M
$0.492 $3
model identifier
google/gemini-3-flash-preview
Wcnbai · Gemini CLI −84%
input$0.082
output$0.492
cache$0.008
Apixly · gemini cli −60%
input$0.2
output$1.2
cache$0.02
Apixly · gemini vertex −50%
input$0.25
output$1.5
cache$0.025
Apixly · image
input$0.5
output$3
cache$0.05
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 2.5 Pro

context 1 M

High-quality reasoning across text, images, and code.

INPUT / 1M
$0.5 $1.25
OUTPUT / 1M
$4 $10
model identifier
google/gemini-2.5-pro
Apixly · gemini cli −60%
input$0.5
output$4
cache$0.05
Wcnbai · Vertex −8%
input$1.48
output$8.86
cache$0.148
Wcnbai · Gemini优质文本 +53%
input$2.46
output$14.77
cache$0.246
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 2.5 Flash

context 1 M

Fast, low-cost multimodal model with a huge context window.

INPUT / 1M
$0.049 $0.3
OUTPUT / 1M
$0.41 $2.5
model identifier
google/gemini-2.5-flash
Wcnbai · Gemini CLI −84%
input$0.049
output$0.41
cache$0.005
Wcnbai · Vertex −70%
input$0.089
output$0.739
cache$0.009
Apixly · gemini cli −60%
input$0.12
output$1
cache$0.012
Wcnbai · Gemini优质文本 −51%
input$0.148
output$1.23
cache$0.015
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 2.5 Flash Lite

context 1 M

A fast, cost-efficient reasoning model for low-latency applications. It excels at quick text generation and high throughput.

INPUT / 1M
$0.03 $0.1
OUTPUT / 1M
$0.118 $0.4
model identifier
google/gemini-2.5-flash-lite
Wcnbai · Vertex −70%
input$0.03
output$0.118
cache$0.003
Apixly · gemini cli −60%
input$0.04
output$0.16
cache$0.004
Wcnbai · Gemini优质文本 −51%
input$0.049
output$0.197
cache$0.005
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 3.8 Flash

context 1 M

Gemini 3.8 Flash is Google's intelligent Flash model, excelling in software engineering, agentic tasks, and multi-step reasoning.

INPUT / 1M
$0.05 $0.75
OUTPUT / 1M
$0.02 $3.75
model identifier
google/gemini-flash-3.8
Claude Hub · direct −98%
input$0.05
output$0.02
cache$0.05
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
G

Gemini 4 Argon

context 1 M

Gemini 4 Argon is a flagship reasoning model excelling at complex software engineering, mathematical, and scientific problem-solving, and autonomous agents.

INPUT / 1M
$0.2 $2
OUTPUT / 1M
$0.2 $10
model identifier
google/gemini-4-argon
Subscriptions · direct −97%
input$0.2
output$0.2
cache$0.2
Wcnbai · Vertex −70%
input$0.591
output$2.95
cache$0.03
Apixly · gemini vertex −50%
input$1
output$5
cache$0.05
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
D

DeepSeek

2
D

DeepSeek V4.1 Flash

context 1 M

A sparse mixture-of-experts model with a Causal Encoder-Decoder architecture. It excels at complex language tasks.

INPUT / 1M
$0.025 $0.018
OUTPUT / 1M
$0.025 $0.684
model identifier
deepseek-v4.1-flash
Claude Hub · direct −93%
input$0.025
output$0.025
cache$0.025
Wellflow · direct −81%
input$0.03
output$0.1
cache$0.03
Apixly · ChineseLLM-stable −10%
input$0.016
output$0.616
cache$0.005
Apixly · ChineseLLM-discount
input$0.018
output$0.684
cache$0.005
Wcnbai · 国产 +40%
input$0.197
output$0.788
cache$0.02
Wcnbai · 国产模型 +52%
input$0.213
output$0.854
cache$0.021
Wcnbai · KIMI +64%
input$0.23
output$0.919
cache$0.023
Wcnbai · GLM +134%
input$0.328
output$1.31
cache$0.033
Wcnbai · DeepSeek +134%
input$0.328
output$1.31
cache$0.033
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
D

DeepSeek V4 Flash

context 1 M

A highly efficient Mixture-of-Experts model for fast inference. It excels at processing long contexts.

INPUT / 1M
$0.04 $0.091
OUTPUT / 1M
$0.04 $0.182
model identifier
deepseek/deepseek-v4-flash
Claude Hub · direct −71%
input$0.04
output$0.04
cache$0.04
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
K

Kimi

4
K

Kimi K3

context 1 M

Kimi K3 is an open-weight multimodal reasoning model adept at complex coding, knowledge work, and long-horizon agentic workflows.

INPUT / 1M
$0.39 $0.304
OUTPUT / 1M
$1.93 $11.59
model identifier
moonshotai/kimi-k3
Wellflow · direct −80%
input$0.39
output$1.93
cache$0.04
Wcnbai · 腾讯云 +66%
input$3.28
output$16.41
cache$0.328
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
K

Kimi K2.7 Code

context 260 K

Kimi K2.7 Code is a coding-focused model designed for reliable, end-to-end programming tasks over long contexts. It utilizes a multimodal mixture-of-experts architecture.

INPUT / 1M
$0.694 $0.671
OUTPUT / 1M
$2.88 $3.35
model identifier
moonshotai/kimi-k2.7-code
Wcnbai · 国产模型 −11%
input$0.694
output$2.88
cache$0.139
Wcnbai · KIMI −4%
input$0.747
output$3.1
cache$0.149
Apixly · ChineseLLM-discount
input$0.671
output$3.35
cache$0.18
Wcnbai · DeepSeek +37%
input$1.07
output$4.43
cache$0.213
Wcnbai · 阿里百炼 +37%
input$1.07
output$4.43
cache$0.213
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
K

Kimi K2.6

context 260 K

Kimi K2.6 is a multimodal model excelling at long-horizon coding, UI/UX generation, and multi-agent orchestration across Python, Rust, and Go.

INPUT / 1M
$0.585 $0.65
OUTPUT / 1M
$3.07 $3.41
model identifier
moonshotai/kimi-k2.6
Apixly · ChineseLLM-stable −10%
input$0.585
output$3.07
cache$0.135
Wcnbai · DeepSeek +35%
input$1.07
output$4.43
cache$0.181
Wcnbai · 阿里百炼 +35%
input$1.07
output$4.43
cache$0.181
Wcnbai · GLM +35%
input$1.07
output$4.43
cache$0.181
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
K

Kimi K2.5

context 260 K

Kimi K2.5 is a multimodal model excelling at visual coding and agent swarm coordination. It leverages extensive pretraining for advanced capabilities.

INPUT / 1M
$0.427 $0.5
OUTPUT / 1M
$2.24 $2.5
model identifier
moonshotai/kimi-k2.5
Wcnbai · 国产模型 −11%
input$0.427
output$2.24
cache—
Wcnbai · 阿里百炼 +37%
input$0.657
output$3.45
cache—
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
GL

GLM

5
GL

GLM 5.3

context 1 M

A large-scale reasoning model for complex software engineering and long-horizon agent tasks. It excels at handling extensive context for intricate problem-solving.

INPUT / 1M
$0.19 $0.079
OUTPUT / 1M
$0.19 $4.8
model identifier
z-ai/glm-5.3
China Subscriptions · direct −92%
input$0.19
output$0.19
cache—
Claude Hub · direct −92%
input$0.2
output$0.2
cache$0.2
Wellflow · direct −90%
input$0.12
output$0.38
cache$0.04
Wcnbai · 国产 −27%
input$0.788
output$2.76
cache$0.197
Wcnbai · 国产模型 −21%
input$0.854
output$2.99
cache$0.213
Wcnbai · Qwen +21%
input$1.31
output$4.6
cache$0.328
Wcnbai · DeepSeek +21%
input$1.31
output$4.6
cache$0.328
Wcnbai · 阿里百炼 +21%
input$1.31
output$4.6
cache$0.328
Wcnbai · 腾讯云 +21%
input$1.31
output$4.6
cache$0.328
Wcnbai · GLM +21%
input$1.31
output$4.6
cache$0.328
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
GL

GLM 5.3 Flash

context 1.3 M

A native multimodal model for efficient coding and long-horizon agent tasks. It excels at maintaining accurate long-context behavior.

INPUT / 1M
$0.06 $0.15
OUTPUT / 1M
$0.2 $0.5
model identifier
z-ai/glm-5.3-flash
Wellflow · direct −60%
input$0.06
output$0.2
cache$0.02
Wcnbai · Qwen −9%
input$0.131
output$0.46
cache$0.038
Wcnbai · GLM −9%
input$0.131
output$0.46
cache$0.038
Wcnbai · 腾讯云 −9%
input$0.131
output$0.46
cache$0.038
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
GL

GLM 5.2

context 1 M

GLM 5.2 is a large-scale reasoning model from Z.ai. It supports text input and output with a 1M-token context window, and is suited for long-horizon agent workflows, project-level software engineering, and complex multi-step automation.

INPUT / 1M
$0.19 $0.62
OUTPUT / 1M
$0.19 $2.42
model identifier
z-ai/glm-5.2
China Subscriptions · direct −88%
input$0.19
output$0.19
cache—
Wcnbai · 国产 +17%
input$0.788
output$2.76
cache$0.197
Wcnbai · 国产模型 +26%
input$0.854
output$2.99
cache$0.213
Wcnbai · KIMI +36%
input$0.919
output$3.22
cache$0.23
Wcnbai · GLM +94%
input$1.31
output$4.6
cache$0.328
Wcnbai · DeepSeek +94%
input$1.31
output$4.6
cache$0.328
Wcnbai · 阿里百炼 +94%
input$1.31
output$4.6
cache$0.328
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
GL

GLM 5.1

context 200 K

GLM-5.1 excels at long-horizon coding tasks, offering independent and continuous operation for complex development.

INPUT / 1M
$0.2 $0.966
OUTPUT / 1M
$0.2 $3.04
model identifier
z-ai/glm-5.1
Wellflow · direct −90%
input$0.2
output$0.2
cache$0.2
Wcnbai · 国产 −11%
input$0.788
output$2.76
cache—
Apixly · ChineseLLM-stable −10%
input$0.869
output$2.73
cache$0.161
Wcnbai · 国产模型 −4%
input$0.854
output$2.99
cache—
Apixly · ChineseLLM-discount
input$0.966
output$3.04
cache$0.179
Wcnbai · KIMI +3%
input$0.919
output$3.22
cache—
Wcnbai · DeepSeek +48%
input$1.31
output$4.6
cache—
Wcnbai · GLM +48%
input$1.31
output$4.6
cache—
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
GL

GLM 5

context 200 K

GLM-5 is an open-source foundation model for complex systems design and long-horizon agent workflows, excelling at large-scale programming tasks.

INPUT / 1M
$1.18 $1
OUTPUT / 1M
$4.33 $3.2
model identifier
z-ai/glm-5
Wcnbai · 国产 +31%
input$1.18
output$4.33
cache—
Wcnbai · 国产模型 +42%
input$1.28
output$4.69
cache—
Wcnbai · DeepSeek +119%
input$1.97
output$7.22
cache—
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
M

MiniMax

2
M

MiniMax M3

context 1 M

This multimodal foundation model accepts text, image, and video inputs to generate text outputs, excelling at long-horizon agentic tasks and coding.

INPUT / 1M
$0.04 $0.3
OUTPUT / 1M
$0.13 $1.2
model identifier
minimax/minimax-m3
Wellflow · direct −89%
input$0.04
output$0.13
cache$0.01
China Subscriptions · direct −60%
input$0.3
output$0.3
cache—
Apixly · ChineseLLM-stable −10%
input$0.27
output$1.08
cache$0.054
Wcnbai · M3 +268%
input$1.1
output$4.41
cache$0.221
Wcnbai · M3-cc接口 +360%
input$1.38
output$5.52
cache$0.276
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
M

MiniMax M2.7

context 200 K

MiniMax-M2.7 is a large language model designed for autonomous productivity and continuous improvement. It excels at active participation in its own evolution through advanced agentic capabilities.

INPUT / 1M
$0.224 $0.3
OUTPUT / 1M
$0.896 $1.2
model identifier
minimax/minimax-m2.7
Wcnbai · 国产模型 −25%
input$0.224
output$0.896
cache$0.045
Wcnbai · minimax-cc接口 −8%
input$0.276
output$1.1
cache$0.055
Wcnbai · minimax −8%
input$0.276
output$1.1
cache$0.055
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
xA

xAI

3
xA

Grok 4.7

context 500 K

Grok 4.7 is a powerful model excelling at complex coding, agentic tasks, and knowledge work, especially for long-running software engineering projects.

INPUT / 1M
$0.18 $2
OUTPUT / 1M
$0.53 $6
model identifier
x-ai/grok-4.7
Wellflow · direct −91%
input$0.18
output$0.53
cache$0.05
Apixly · gpt −90%
input$0.2
output$0.6
cache$0.05
Wcnbai · grok-官转 −74%
input$0.525
output$1.58
cache$0.131
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
xA

Grok 4.6

context 500 K

Grok 4.6 is SpaceXAI's most advanced model, excelling in coding, knowledge work, and STEM tasks.

INPUT / 1M
$0.08 $2
OUTPUT / 1M
$0.08 $6
model identifier
x-ai/grok-4.6
Claude Hub · direct −98%
input$0.08
output$0.08
cache$0.08
Apixly · gpt −90%
input$0.2
output$0.6
cache$0.05
Wcnbai · grok-官转 −74%
input$0.525
output$1.58
cache$0.131
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
xA

Grok 4.5

context 500 K

Grok 4.5 is SpaceXAI's most advanced model, excelling in coding, knowledge work, and STEM tasks.

INPUT / 1M
$0.18 $2
OUTPUT / 1M
$0.54 $6
model identifier
x-ai/grok-4.5
Wellflow · direct −91%
input$0.18
output$0.54
cache$0.03
Apixly · gpt −90%
input$0.2
output$0.6
cache$0.03
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
Q

Qwen

7
Q

Qwen3.8 Flash

context 1 M

Qwen3.8 Flash is a multimodal reasoning model for coding, agentic workflows, visual understanding, and document/codebase analysis. It excels at desktop interaction and analyzing charts and long videos.

INPUT / 1M
$0.03 $0.15
OUTPUT / 1M
$0.07 $0.47
model identifier
qwen/qwen3.8-flash
Wellflow · direct −84%
input$0.03
output$0.07
cache$0.03
Wcnbai · 国产 −44%
input$0.079
output$0.266
cache$0.016
Wcnbai · 国产模型 −40%
input$0.085
output$0.288
cache$0.017
Wcnbai · KIMI −35%
input$0.092
output$0.31
cache$0.018
Apixly · ChineseLLM-stable −10%
input$0.135
output$0.423
cache$0.014
Wcnbai · DeepSeek −7%
input$0.131
output$0.443
cache$0.026
Wcnbai · GLM −7%
input$0.131
output$0.443
cache$0.026
Wcnbai · Qwen −7%
input$0.131
output$0.443
cache$0.026
Apixly · ChineseLLM-discount
input$0.15
output$0.47
cache$0.016
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
Q

Qwen3.7 Max

context 1 M

Qwen3.7-Max is a flagship model for agent-centric workloads, excelling at coding and office productivity tasks.

INPUT / 1M
$0.25 $1.48
OUTPUT / 1M
$0.25 $4.42
model identifier
qwen/qwen3.7-max
Wellflow · direct −92%
input$0.25
output$0.25
cache$0.25
Apixly · ChineseLLM-stable −10%
input$1.33
output$3.98
cache$0.266
Wcnbai · Qwen +34%
input$1.97
output$5.91
cache$0.394
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
Q

Qwen3.7 Plus

context 1 M

A multimodal model that excels at text generation and understanding, with enhanced capabilities for image processing.

INPUT / 1M
$0.09 $0.32
OUTPUT / 1M
$0.09 $1.28
model identifier
qwen/qwen3.7-plus
China Subscriptions · direct −89%
input$0.09
output$0.09
cache—
Wellflow · direct −81%
input$0.15
output$0.15
cache$0.15
Apixly · ChineseLLM-stable −10%
input$0.288
output$1.15
cache$0.058
Wcnbai · 国产模型 +300%
input$1.28
output$5.12
cache$0.085
Wcnbai · 阿里百炼 +516%
input$1.97
output$7.88
cache$0.131
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
Q

Qwen3.6 Flash

context 1 M

Qwen3.6 Flash is a fast, efficient multimodal model supporting text, image, and video input. It excels at processing long contexts.

INPUT / 1M
$0.197 $0.188
OUTPUT / 1M
$1.18 $1.12
model identifier
qwen/qwen3.6-flash
Wcnbai · 阿里百炼 +5%
input$0.197
output$1.18
cache—
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
Q

Qwen3.5-Flash

context 1 M

Qwen3.5 Flash models are efficient vision-language models with a hybrid architecture, excelling at fast inference for multimodal tasks.

INPUT / 1M
$0.1 $0.065
OUTPUT / 1M
$0.1 $0.26
model identifier
qwen/qwen3.5-flash
Wellflow · direct −38%
input$0.1
output$0.1
cache$0.1
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
Q

Qwen3.6 Plus

context 1 M

Qwen 3.6 Plus is a powerful, scalable LLM with a hybrid architecture, excelling at high-performance inference.

INPUT / 1M
$0.1 $0.325
OUTPUT / 1M
$0.1 $1.95
model identifier
qwen/qwen3.6-plus
China Subscriptions · direct −91%
input$0.1
output$0.1
cache—
Wcnbai · 国产 −39%
input$0.197
output$1.18
cache$0.02
Wcnbai · 国产模型 −34%
input$0.213
output$1.28
cache$0.021
Wcnbai · 阿里百炼 +1%
input$0.328
output$1.97
cache$0.033
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure
Q

Qwen3.8 Omni Flash

context 1 M

Qwen3.8 Omni Flash is an omni-modal reasoning model with native audio-video understanding, excelling at audio-video analysis and summarization.

INPUT / 1M
$0.03 $0.15
OUTPUT / 1M
$0.08 $0.47
model identifier
qwen/qwen3.8-omni-flash
Wellflow · direct −82%
input$0.03
output$0.08
cache$0.01
Discount shown against the model's official price · prices per 1M tokens · DarkRouter takes the cheapest working route and switches over on failure