AllToken/Model Rankings·7 Days

Decisions, quantified.

Top models ranked by request volume and token usage

Model Leaderboard

Ranked by routed token usage in the last 7 Days.
01
DeepSeek V4 Flash
deepseek-v4-flash
deepseek
↑ 87.0%
View →
02
GLM 5.3 Flash
glm-5.3-flash
bigmodel
· 0.0%
View →
03
GPT-5.6 Luna
gpt-5.6-luna
openai
↑ 20.0%
View →
04
DeepSeek V4 Pro
deepseek-v4-pro
deepseek
↑ 42266608.0%
View →
05
Claude Sonnet 4.6
claude-sonnet-4-6
anthropic
↑ 30942188.0%
View →
06
Claude Opus 4.8
claude-opus-4-8
anthropic
↑ 927506.0%
View →
07
Claude Opus 4.6
claude-opus-4-6
anthropic
↑ 1375729.0%
View →
08
DeepSeek V4 Flash Vision (Exp)
deepseek-v4-flash-vision-exp
deepseek
↑ 736657.0%
View →
09
Claude Opus 4.7
claude-opus-4-7
anthropic
↑ 943128.0%
View →
10
GLM 4.7 Flash
glm-4.7-flash
bigmodel
↑ 26052.0%
View →

Market Share

Provider share across routed token usage.
deepseekTop model: DeepSeek V4 Flash47.0%
openaiTop model: GPT-4o Mini20.2%
bigmodelTop model: GLM 5.3 Flash15.3%
anthropicTop model: Claude Sonnet 4.65.8%
sufyTop model: GPT-5.6 Luna5.2%
tencent-tokenhub-sgpTop model: —4.9%
aliyun-dashscopeTop model: Kimi K31.2%
tencent-tokenhubTop model: GLM 5.20.3%
google-geminiTop model: Gemini 3 Pro Image (Nano Banana Pro)0.0%
moonshot-kimiTop model: Kimi K2.7 Code0.0%

Benchmarks

Intelligence Index

Intelligence

Composite benchmark index

01Claude Opus 4.6anthropic96.8
02Claude Sonnet 4.6anthropic94.5
03MiniMax M2.7Unknown92.3
04MiniMax M2.5Unknown89.7
05Claude Haiku 4.5anthropic87.2
06Seed 2.0 LiteUnknown85.1
07MiniMax M2.7 Highspeedminimax-official84.6
08MiniMax M2.5 Highspeedminimax-official83.9

Coding

Agentic coding score

01Claude Sonnet 4.6anthropic94.1
02Claude Opus 4.6anthropic93.5
03MiniMax M2.7Unknown88.6
04MiniMax M2.5Unknown85.4
05Claude Haiku 4.5anthropic82.3
06Seed 2.0 LiteUnknown78.9

Math

Mathematical reasoning

01Claude Opus 4.6anthropic97.3
02MiniMax M2.7Unknown94.8
03Claude Sonnet 4.6anthropic93.6
04MiniMax M2.5Unknown90.2
05Claude Haiku 4.5anthropic85.5
06Seed 2.0 LiteUnknown82.7

Knowledge

MMLU benchmark score

01Claude Opus 4.6anthropic95.2
02Claude Sonnet 4.6anthropic93.8
03MiniMax M2.7Unknown91.5
04MiniMax M2.5Unknown88.3
05Claude Haiku 4.5anthropic86.7
06Seed 2.0 LiteUnknown84.2

Fastest Models

Highest Throughput
ModelHighest ThroughputLowest LatencyTime to First Token
MiniMax M2.7 Highspeed
minimax-official
312.5 t/s1520 ms350 ms
MiniMax M2.5 Highspeed
minimax-official
298.8 t/s1280 ms320 ms
Claude Haiku 4.5
anthropic
245.6 t/s850 ms180 ms
MiniMax M2.7
Unknown
178.3 t/s2350 ms560 ms
MiniMax M2.5
Unknown
165.4 t/s1780 ms480 ms
Claude Sonnet 4.6
anthropic
142.8 t/s1450 ms420 ms
Seed 2.0 Lite
Unknown
118.5 t/s3200 ms920 ms
Claude Opus 4.6
anthropic
86.2 t/s2100 ms680 ms
MiniMax M2.7 Highspeed
minimax-official· 312.5 t/s· 1520 ms
MiniMax M2.5 Highspeed
minimax-official· 298.8 t/s· 1280 ms
Claude Haiku 4.5
anthropic· 245.6 t/s· 850 ms
MiniMax M2.7
Unknown· 178.3 t/s· 2350 ms
MiniMax M2.5
Unknown· 165.4 t/s· 1780 ms
Claude Sonnet 4.6
anthropic· 142.8 t/s· 1450 ms
Seed 2.0 Lite
Unknown· 118.5 t/s· 3200 ms
Claude Opus 4.6
anthropic· 86.2 t/s· 2100 ms

Categories

Ranked leaders by use case.
01
DeepSeek V4 Flash
deepseek-v4-flash
deepseek
↑ 87.0%
View →
02
GLM 5.3 Flash
glm-5.3-flash
bigmodel
· 0.0%
View →
03
GPT-5.6 Luna
gpt-5.6-luna
openai
↑ 20.0%
View →
04
DeepSeek V4 Pro
deepseek-v4-pro
deepseek
↑ 42266608.0%
View →
05
Claude Sonnet 4.6
claude-sonnet-4-6
anthropic
↑ 30942188.0%
View →
06
Claude Opus 4.8
claude-opus-4-8
anthropic
↑ 927506.0%
View →
07
Claude Opus 4.6
claude-opus-4-6
anthropic
↑ 1375729.0%
View →
08
DeepSeek V4 Flash Vision (Exp)
deepseek-v4-flash-vision-exp
deepseek
↑ 736657.0%
View →
09
Claude Opus 4.7
claude-opus-4-7
anthropic
↑ 943128.0%
View →
10
GLM 4.7 Flash
glm-4.7-flash
bigmodel
↑ 26052.0%
View →