Home › Rankings

Model Rankings

The Bridge catalog ranked by developer demand — what teams actually build on when they switch to Chinese frontier models.

Most popular

1

DeepSeek V4 Pro

DeepSeek · reasoning & coding flagship

The default pick for hard reasoning and code. Flat pricing with no peak-hour surcharges. $0.28 / $0.82 $1.27 / $3.80save 78%

reasoningcodegeneral
2

GLM-5.2

Zhipu AI · Chinese language + vision

The strongest Chinese-language understanding of the major labs, with native vision input. $0.56 / $1.41 $1.13 / $3.94save 50%

visionchinesemultimodal
3

DeepSeek V4 Flash

DeepSeek · high-volume value tier

The best volume-per-dollar in the DeepSeek family for production pipelines. $0.14 / $0.28 $0.42 / $1.27save 67%

fasthigh-volume
4

Kimi K3

Moonshot AI · long-context agents

256K context for whole-repo review, contract analysis and agentic loops. $2.47 / $12.34 $3.08 / $15.42save 20%

long-contextagent
5

Qwen3.7-Max

Alibaba Cloud · flagship reasoning

Qwen's flagship reasoning tier, strong across languages. $1.01 / $3.04 $1.69 / $5.07save 40%

flagshipreasoning
Rankings reflect developer demand observed on the platform and in the market. As Bridge's traffic grows, this page will publish token-based leaderboards (by week and month) computed from real, aggregated usage — in the meantime this list is kept current with the live catalog on the models page.

New this month

Kimi K2.6

Moonshot AI · long-context generalist

256K context at a fraction of flagship cost — the agent workhorse. $0.56 / $1.13

Kimi K2.7 Code

Moonshot AI · coding specialist

Code-tuned reasoning for agents and IDE backends. $0.56 / $1.13

MiniMax M3

MiniMax · newest flagship

The latest MiniMax generation at value pricing. $0.28 / $0.59

GLM-5.3

Zhipu AI · latest flagship

Zhipu's newest flagship — stronger reasoning, 30% below official. $1.01 / $3.17 $1.44 / $4.52save 30%

MiniMax M2.7

MiniMax · all-round assistant

Strong general performance at competitive flat pricing. $0.30 / $1.18

Best value for production

$

Qwen2.5 72B Instruct

Open weights · self-hosted

The lowest p50 latency on Bridge, for high-volume classification, extraction and translation. $0.25 / $0.75

$

Qwen3 Embedding 8B

Open weights · embeddings

Near-zero-cost embeddings for RAG and semantic search. $0.02 per 1M tokens

Get $1 free credit → try every ranked model