Submit

Chinese LLM leaderboards

Composite "best for X" tables ranked from live catalog data, plus per-benchmark top 10s at the bottom. Only official and auditable third-party benchmark scores — no in-house evals.

Cheapest Chinese chat models

Lowest blended price / 1M tokens for models tagged chat or multilingual. Picked the cheapest hosting per model.

#ModelCheapest / 1MViaContext
1Doubao Lite 32K
ByteDance
$0.05ByteDance Doubao (Volcengine)33K
2GLM-4-Air
Zhipu AI
$0.07Zhipu AI131K
3Yi Lightning
01.AI
$0.1401.AI16K
4Step 3.5 Flash
StepFun (阶跃星辰) · open-weight
$0.15StepFun262K
5Doubao Pro 32K
ByteDance
$0.15ByteDance Doubao (Volcengine)33K
6DeepSeek V4 Flash
DeepSeek · open-weight
$0.17DeepSeek1049K
7DeepSeek V3.1
DeepSeek · open-weight
$0.30DeepSeek33K
8DeepSeek V3
DeepSeek AI · open-weight
$0.48DeepSeek66K
9MiniMax M2.7
MiniMax · open-weight
$0.52MiniMax197K
10DeepSeek V4 Pro
DeepSeek · open-weight
$0.54DeepSeek1049K

Cheapest Chinese reasoning models

Lowest blended price for reasoning / chain-of-thought / tool-use models. Same pricing methodology as the chat board.

#ModelCheapest / 1MViaContext
1GLM-4-Air
Zhipu AI
$0.07Zhipu AI131K
2Yi Lightning
01.AI
$0.1401.AI16K
3Step 3.5 Flash
StepFun (阶跃星辰) · open-weight
$0.15StepFun262K
4Doubao Pro 32K
ByteDance
$0.15ByteDance Doubao (Volcengine)33K
5DeepSeek V3.1
DeepSeek · open-weight
$0.30DeepSeek33K
6DeepSeek V3
DeepSeek AI · open-weight
$0.48DeepSeek66K
7DeepSeek V4 Pro
DeepSeek · open-weight
$0.54DeepSeek1049K
8Qwen3.5-122B-A10B
Alibaba Cloud · open-weight
$0.71Alibaba Cloud DashScope262K
9Kimi K2.5
Moonshot AI · open-weight
$0.80Moonshot AI262K
10Qwen3.5 397B A17B
Alibaba Cloud · open-weight
$0.88Alibaba Cloud DashScope262K

Longest context windows

Chat / reasoning / multimodal models ranked by context window. For document Q&A, whole-repo code review, or long transcripts. Embedding / image-gen / video types excluded — their 'context' is a different quantity.

#ModelCheapest / 1MViaContext
1DeepSeek V4 Flash
DeepSeek · open-weight
$0.17DeepSeek1049K
2DeepSeek V4 Pro
DeepSeek · open-weight
$0.54DeepSeek1049K
3MiniMax-01
MiniMax · open-weight
$0.42MiniMax1000K
4Qwen3.5 Plus 2026-04-20
Alibaba Cloud
$0.90Alibaba Cloud DashScope1000K
5Qwen3.6 Flash
Alibaba Cloud
$0.56Alibaba Cloud DashScope1000K
6Qwen3.6 Plus
Alibaba Cloud · open-weight
$0.73Alibaba Cloud DashScope1000K
7Qwen3.5-Flash
Alibaba Cloud
$0.11Alibaba Cloud DashScope1000K
8Qwen3.5 Plus 2026-02-15
Alibaba Cloud · open-weight
$0.58Alibaba Cloud DashScope1000K
9Qwen3 Coder Plus
Alibaba Cloud · open-weight
$1.30Alibaba Cloud DashScope1000K
10Qwen3 Coder Flash
Alibaba Cloud · open-weight
$0.39Alibaba Cloud DashScope1000K

Open-weight Chinese models

Models that ship weights you can self-host. Ordered by context window — biggest first.

#ModelCheapest / 1MViaContext
1DeepSeek V4 Flash
DeepSeek · open-weight
$0.17DeepSeek1049K
2DeepSeek V4 Pro
DeepSeek · open-weight
$0.54DeepSeek1049K
3MiniMax-01
MiniMax · open-weight
$0.42MiniMax1000K
4Qwen3.6 Plus
Alibaba Cloud · open-weight
$0.73Alibaba Cloud DashScope1000K
5Qwen3.5 Plus 2026-02-15
Alibaba Cloud · open-weight
$0.58Alibaba Cloud DashScope1000K
6Qwen3 Coder Plus
Alibaba Cloud · open-weight
$1.30Alibaba Cloud DashScope1000K
7Qwen3 Coder Flash
Alibaba Cloud · open-weight
$0.39Alibaba Cloud DashScope1000K
8Qwen Plus 0728 (thinking)
Alibaba Cloud · open-weight
$0.39Alibaba Cloud DashScope1000K
9Qwen Plus 0728
Alibaba Cloud · open-weight
$0.39Alibaba Cloud DashScope1000K
10MiniMax M1
MiniMax · open-weight
$0.85MiniMax1000K

Best Chinese coding models

Filter: `code` capability. Ranked by top HumanEval / LiveCodeBench / SWE-bench score when available — otherwise the model is hidden to avoid a misleading placeholder.

#ModelCheapest / 1MViaContextScore
1R1
DeepSeek · open-weight
$1.15DeepSeek64K89.1%
HumanEval
2Llama 3.3 70B Instruct
Meta · open-weight
$0.88Together.ai131K88.4%
HumanEval
3DeepSeek V3
DeepSeek AI · open-weight
$0.48DeepSeek66K82.5%
HumanEval

China-only hosted models

Models whose every published hosting sits on a Provider without an overseas node. For CN-mainland users who don't want cross-border egress.

#ModelCheapest / 1MViaContext
1GLM-4 Plus
Zhipu AI
$0.00Zhipu AI128K
2GLM 4.5 Air (free)
Zhipu AI · open-weight
$0.00Zhipu AI131K
3Doubao Lite 32K
ByteDance
$0.05ByteDance Doubao (Volcengine)33K
4GLM-4-Air
Zhipu AI
$0.07Zhipu AI131K
5GLM 4 32B
Zhipu AI · open-weight
$0.10Zhipu AI128K
6Hy3 preview
Tencent · open-weight
$0.11Tencent Hunyuan262K
7UI-TARS 7B
ByteDance · open-weight
$0.13ByteDance Doubao (Volcengine)128K
8Seed 1.6 Flash
ByteDance · open-weight
$0.13ByteDance Doubao (Volcengine)262K
9GLM 4.7 Flash
Zhipu AI · open-weight
$0.14Zhipu AI203K
10Step 3.5 Flash
StepFun (阶跃星辰) · open-weight
$0.15StepFun262K

Per-benchmark top 10

Existing published benchmark scores — sourced from papers or auditable third-party boards. Click any model to see its full profile.

HumanEval coding

Higher = better
#ModelScoreConditionsSource
1Hunyuan-Large
Tencent
90.00instruct modelofficial
2Hunyuan-Large
Tencent
90.00instructofficial
3MiniMax-Text-01
MiniMax
86.90—official
4Hunyuan-Large
Tencent
71.40pre-trained modelofficial
5Hunyuan-Large
Tencent
71.40pretrainedofficial
6R1
DeepSeek
0.890-shot pass@1official
7Llama 3.3 70B Instruct
Meta
0.880-shot pass@1official
8DeepSeek V3
DeepSeek AI
0.820-shot pass@1official
9Qwen3 72B Instruct
Alibaba
0.820-shot pass@1official

MMLU knowledge

Higher = better
#ModelScoreConditionsSource
1Hunyuan-Large
Tencent
89.90instruct modelofficial
2Hunyuan-Large
Tencent
89.90instructofficial
3MiniMax-Text-01
MiniMax
88.50—official
4Hunyuan-Large
Tencent
88.40pre-trained modelofficial
5Hunyuan-Large
Tencent
88.40pretrainedofficial
6R1
DeepSeek
0.905-shotofficial
7DeepSeek V3
DeepSeek AI
0.895-shotofficial
8Llama 3.3 70B Instruct
Meta
0.875-shotofficial
9Qwen3 72B Instruct
Alibaba
0.865-shotofficial