> model race

Who's winning the model race?

A twice-daily Boostor snapshot for frontier model performance, coding strength, agentic readiness, tool support, and cost efficiency.

tooling feed live, benchmark key neededrefresh every 12h/api/model-race

leader

models

68%

data health

Jul 7

updated

> leaderboard

Composite race score

Sort by any column, filter by provider, and refresh live data. Artificial Analysis drives the benchmark score when configured. Without that key, the page clearly falls back to OpenRouter's live tooling and economics feed.

JSON feed

sort

provider

DeepSeek V4 Flash

tooling proxy

DeepSeek · 1M ctx

95.1

intel

n/a

code

n/a

agent

n/a

tools

83.3

value

98.5

source

MiMo-V2.5

tooling proxy

Xiaomi · 1M ctx

93.7

intel

n/a

code

n/a

agent

n/a

tools

83.3

value

97.8

source

MiniMax M3

tooling proxy

MiniMax · 1M ctx

91.7

intel

n/a

code

n/a

agent

n/a

tools

83.3

value

93.1

source

GLM 5.2

tooling proxy

Z.ai · 1M ctx

89.5

intel

n/a

code

n/a

agent

n/a

tools

83.3

value

87.3

source

DeepSeek V4 Pro

tooling proxy

DeepSeek · 1M ctx

88.1

intel

n/a

code

n/a

agent

n/a

tools

83.3

value

source

Claude Opus 4.7

tooling proxy

Anthropic · 1M ctx

82.5

intel

n/a

code

n/a

agent

n/a

tools

83.3

value

67.4

source

Step 3.7 Flash

tooling proxy

StepFun · 256K ctx

82.2

intel

n/a

code

n/a

agent

n/a

tools

83.3

value

93.6

source

Claude Opus 4.8

tooling proxy

Anthropic · 1M ctx

81.2

intel

n/a

code

n/a

agent

n/a

tools

83.3

value

67.4

source

Claude Sonnet 4.6

tooling proxy

Anthropic · 1M ctx

79.5

intel

n/a

code

n/a

agent

n/a

tools

83.3

value

72.7

source

Hy3 preview

tooling proxy

Tencent · 262K ctx

78.7

intel

n/a

code

n/a

agent

n/a

tools

value

98.4

source

Gemini 3 Flash Preview

tooling proxy

Google · 1M ctx

78.2

intel

n/a

code

n/a

agent

n/a

tools

83.3

value

87.6

source

GPT-5.5

tooling proxy

OpenAI · 1.1M ctx

77.2

intel

n/a

code

n/a

agent

n/a

tools

83.3

value

65.6

source

> subcharts

Tooling and value signals

Benchmarks pick winners. Builders also need to know which model can call tools, return structured output, fit the repo, and stay affordable.

Rank my stack

Tool-calling readiness

OpenRouter-supported parameters, filtered to agent-builder relevant capabilities.

subchart

model	tools	structured outputs	reasoning	response format	web search	score
DeepSeek V4 Flash DeepSeek

model

tools

structured outputs

reasoning

response format

web search

score

DeepSeek V4 Flash

DeepSeek

> sources

Benchmark coverage map

The page separates structured live feeds from tracked benchmark sources. That keeps the chart useful without pretending every leaderboard exposes a clean API.

Who's winning the model race?

Composite race score

DeepSeek V4 Flash

MiMo-V2.5

MiniMax M3

GLM 5.2

DeepSeek V4 Pro

Claude Opus 4.7

Step 3.7 Flash

Claude Opus 4.8

Claude Sonnet 4.6

Hy3 preview

Gemini 3 Flash Preview

GPT-5.5

Tooling and value signals

Tool-calling readiness

Benchmark coverage map

DeepSeek V4 Flash

tooling feed live, benchmark key needed

Price, score, context