Skip to main content

BenchLM recommendation

Best Chinese AI Models in 2026

Data verified

Kimi K3 leads chinese ai models on BenchLM's July 2026 rankings with a score of 81, ahead of Qwen3.7 Max (72.8) and MiMo-V2.5-Pro (70.2). Each row shows its evidence status and conditional 90% score interval.

Last verified: July 20, 2026

Use this page for the current Chinese-model order. The table rebuilds from the active public ranking lane rather than preserving a copied winner, so the first row changes with the same contract used by the overall leaderboard.

This is the public BenchAlign v5 overall lane filtered to tracked Chinese labs. “Chinese” describes lab origin here; it is not a score of Chinese-language quality.

Bottom line: start with the live overall leader when broad benchmark performance is the constraint. Switch to the highest open-weight row when deployment control matters, and do not use a narrow point gap as the only tie-breaker.

The live Chinese-lab slice currently starts with Kimi K3, followed by Qwen3.7 Max and MiMo-V2.5-Pro. All three use the same BenchAlign v5 contract as the overall leaderboard. Every row below shows its evidence label and conditional 90% score interval because narrow point gaps should not look decisive.

The best open-weight option is MiniMax M3 (ranked #4 with a score of 69.8). While proprietary models lead, open-weight options are within striking distance for teams willing to trade a few points of performance for full model control.

This ranking is based on the public BenchAlign v5 overall contract tracked by BenchLM.ai. For detailed model profiles, click any model name below. To compare two specific models head-to-head, use the "vs #" links.

How to choose

Full Rankings (69 models)

1
Kimi K3
Moonshot AI·Pending·1.05M

81

BenchAlign v5

Supported

90% interval 77.87–84.04

2
Qwen3.7 Max
Alibaba·Proprietary·1M

72.8

BenchAlign v5

Supported

90% interval 66.59–79.08

3
MiMo-V2.5-Pro
Xiaomi·Proprietary·1M

70.2

BenchAlign v5

Supported

90% interval 62.93–77.44

4
MiniMax M3
MiniMax·Open Weight·1M

69.8

BenchAlign v5

Supported

90% interval 65.53–73.98

5
MiMo-V2-Pro
Xiaomi·Proprietary·1M

67.8

BenchAlign v5

Supported

90% interval 60.49–75.06

6
GLM-5.1
Z.AI·Open Weight·203K

67.7

BenchAlign v5

Supported

90% interval 58.01–77.46

7
Qwen3.7 Plus
Alibaba·Proprietary·1M

67.2

BenchAlign v5

Supported

90% interval 58.08–76.36

8
GLM-5-Turbo
Z.AI·Proprietary·200K

66.9

BenchAlign v5

Supported

90% interval 57.48–76.31

9
GLM-5
Z.AI·Open Weight·200K

66.1

BenchAlign v5

Supported

90% interval 55.41–76.71

10
Qwen3.6 Plus
Alibaba·Proprietary·1M

65.2

BenchAlign v5

Supported

90% interval 56.68–73.71

11
MiniMax M2.7
MiniMax·Open Weight·200K

64.1

BenchAlign v5

Supported

90% interval 57.83–70.40

12
GLM-5.2
Z.AI·Open Weight·1M

64

BenchAlign v5

Estimated

90% interval 48.35–79.58

13
GLM-5V-Turbo
Z.AI·Proprietary·200K

63.5

BenchAlign v5

Supported

90% interval 52.89–74.10

14
MiMo-V2-Omni
Xiaomi·Proprietary·262K

63.2

BenchAlign v5

Supported

90% interval 53.64–72.65

15
GLM-4.7
Z.AI·Open Weight·200K

61.2

BenchAlign v5

Supported

90% interval 47.69–74.62

16
Qwen3.5-27B
Alibaba·Open Weight·262K

60.7

BenchAlign v5

Supported

90% interval 52.05–69.35

17
DeepSeek V4 Pro
DeepSeek·Open Weight·1M

60.7

BenchAlign v5

Supported

90% interval 42.29–79.04

18
Qwen3.5-122B-A10B
Alibaba·Open Weight·262K

60.6

BenchAlign v5

Supported

90% interval 50.16–70.96

19
GLM-5 (Reasoning)
Z.AI·Open Weight·200K

59.8

BenchAlign v5

Estimated

90% interval 48.26–71.29

20
Qwen 3.6 Max (preview)
Alibaba·Proprietary·256K

59.7

BenchAlign v5

Supported

90% interval 40.87–78.57

21
Kimi K2.5
Moonshot AI·Open Weight·256K

59.7

BenchAlign v5

Supported

90% interval 52.50–66.83

22
MiniMax M2.5
MiniMax·Proprietary·128K

59.5

BenchAlign v5

Supported

90% interval 52.22–66.83

23
Qwen3.5 397B (Reasoning)
Alibaba·Open Weight·128K

59.5

BenchAlign v5

Estimated

90% interval 47.98–71.01

24
Kimi K2.5 (Reasoning)
Moonshot AI·Proprietary·128K

59.4

BenchAlign v5

Estimated

90% interval 47.83–70.86

25
DeepSeek V4 Flash
DeepSeek·Open Weight·1M

58.9

BenchAlign v5

Estimated

90% interval 47.36–70.39

26
MiMo-V2.5
Xiaomi·Proprietary·1M

58.6

BenchAlign v5

Estimated

90% interval 47.11–70.14

27
DeepSeek V3.2 (Thinking)
DeepSeek·Open Weight·128K

58.2

BenchAlign v5

Estimated

90% interval 46.64–69.66

28
Qwen3 235B 2507 (Reasoning)
Alibaba·Open Weight·128K

58

BenchAlign v5

Estimated

90% interval 46.50–69.53

29
GLM-4.5
Z.AI·Proprietary·128K

57.6

BenchAlign v5

Estimated

90% interval 46.04–69.07

30
Qwen3.5 397B
Alibaba·Open Weight·128K

57

BenchAlign v5

Estimated

90% interval 45.50–68.52

31
Qwen3.5-35B-A3B
Alibaba·Open Weight·262K

57

BenchAlign v5

Supported

90% interval 46.19–67.75

32
Kimi K2.6
Moonshot AI·Open Weight·256K

56.8

BenchAlign v5

Estimated

90% interval 46.92–66.66

33
Qwen3 235B 2507
Alibaba·Open Weight·128K

56

BenchAlign v5

Estimated

90% interval 44.50–67.53

34
Hy3
Tencent·Open Weight·256K

55.6

BenchAlign v5

Estimated

90% interval 44.13–67.16

35
DeepSeek V4 Pro (High)
DeepSeek·Open Weight·1M

55.5

BenchAlign v5

Estimated

90% interval 43.95–66.98

36
DeepSeek V3.2
DeepSeek·Open Weight·128K

55.4

BenchAlign v5

Supported

90% interval 38.94–71.86

37
GLM-4.6
Z.AI·Open Weight·200K

55.1

BenchAlign v5

Supported

90% interval 37.46–72.79

38
Step 3.5 Flash
StepFun·Open Weight·256K

55.1

BenchAlign v5

Supported

90% interval 42.05–68.16

39
Kimi K2.7 Code
Moonshot AI·Open Weight·256K

55

BenchAlign v5

Estimated

90% interval 42.87–67.12

40
DeepSeek LLM 2.0
DeepSeek·Open Weight·128K

54.5

BenchAlign v5

Estimated

90% interval 43.01–66.04

41
MiMo-V2-Flash
Xiaomi·Open Weight·256K

54.1

BenchAlign v5

Supported

90% interval 40.17–67.95

42
DeepSeek V4 Flash (High)
DeepSeek·Open Weight·1M

54

BenchAlign v5

Estimated

90% interval 42.44–65.47

43
Qwen3.6-27B
Alibaba·Open Weight·262K

53.8

BenchAlign v5

Estimated

90% interval 42.30–65.33

44
DeepSeek V3.1
DeepSeek·Open Weight·128K

53.6

BenchAlign v5

Supported

90% interval 35.22–72.05

45
DeepSeek V3.1 (Reasoning)
DeepSeek·Open Weight·128K

53.4

BenchAlign v5

Supported

90% interval 34.72–72.14

46
Qwen2.5-72B
Alibaba·Open Weight·128K

52.2

BenchAlign v5

Estimated

90% interval 40.64–63.67

47
DeepSeek-R1
DeepSeek·Open Weight·128K

51.7

BenchAlign v5

Supported

90% interval 34.16–69.17

48
Qwen3.6-35B-A3B
Alibaba·Open Weight·262K

51.5

BenchAlign v5

Estimated

90% interval 39.95–62.98

49
GLM-4.7-Flash
Z.AI·Open Weight·200K

51.3

BenchAlign v5

Supported

90% interval 38.20–64.30

50
Step 3.7 Flash
StepFun·Open Weight·256K

50.9

BenchAlign v5

Estimated

90% interval 39.35–62.38

51
DeepSeek Coder 2.0
DeepSeek·Open Weight·128K

50.3

BenchAlign v5

Estimated

90% interval 38.77–61.79

52
Seed 1.6
ByteDance·Proprietary·256K

50.2

BenchAlign v5

Estimated

90% interval 38.72–61.74

53
DeepSeekMath V2
DeepSeek·Open Weight·128K

49.9

BenchAlign v5

Estimated

90% interval 38.37–61.40

54
Qwen2.5-1M
Alibaba·Open Weight·1M

49.9

BenchAlign v5

Estimated

90% interval 38.37–61.40

55
Seed-2.0-Lite
ByteDance·Proprietary·256K

49.8

BenchAlign v5

Estimated

90% interval 38.32–61.35

56
Qwen3 Max
Alibaba·Proprietary·1M

48.2

BenchAlign v5

Estimated

90% interval 36.64–59.67

57
GLM-4.5-Air
Z.AI·Proprietary·128K

47.7

BenchAlign v5

Supported

90% interval 29.71–65.69

58
Qwen3.5 Flash
Alibaba·Proprietary·1M

47.7

BenchAlign v5

Supported

90% interval 24.83–70.47

59
Qwen3.5 Plus
Alibaba·Proprietary·1M

47.2

BenchAlign v5

Estimated

90% interval 33.40–61.01

60
Seed 1.6 Flash
ByteDance·Proprietary·256K

45.1

BenchAlign v5

Estimated

90% interval 33.57–56.60

61
DeepSeek V3
DeepSeek·Open Weight·128K

45

BenchAlign v5

Supported

90% interval 26.51–63.43

62
Moonshot v1
Moonshot AI·Proprietary·128K

44.8

BenchAlign v5

Estimated

90% interval 33.28–56.31

63
Seed-2.0-Mini
ByteDance·Proprietary·256K

44.6

BenchAlign v5

Estimated

90% interval 33.08–56.11

64
Ling 2.6 Flash
InclusionAI·Open Weight·262K

43.9

BenchAlign v5

Estimated

90% interval 32.35–55.38

65
DeepSeek R1 Distill Qwen 32B
DeepSeek·Open Weight·128K

42.6

BenchAlign v5

Estimated

90% interval 31.07–54.09

66
Qwen2.5-VL-32B
Alibaba·Open Weight·32K

39.9

BenchAlign v5

Estimated

90% interval 28.34–51.37

67
Qwen2.5 Coder 32B Instruct
Alibaba·Open Weight·128K

34.7

BenchAlign v5

Supported

90% interval 18.46–50.94

68
Kimi K2
Moonshot AI·Proprietary·128K

27.2

BenchAlign v5

Supported

90% interval 16.38–37.99

69
MiniMax M1 80k
MiniMax·Proprietary·80K

25.1

BenchAlign v5

Supported

90% interval 14.66–35.59

Key Takeaways

The top model is Kimi K3 by Moonshot AI with a BenchAlign v5 score of 81 and Supported evidence.

The best open-weight model is MiniMax M3 at position #4.

69 models are included in this ranking.

Score in Context

What these scores mean

Rows use the same current public overall contract as the main leaderboard. On BenchAlign v5 builds, Supported and Estimated describe the evidence behind a position; they are not separate score scales.

Known limitations

Lab origin is a catalog classification, not a Chinese-language evaluation. The ranking does not score license terms, serving cost, throughput, data residency, regional API availability, or fit for a private workload.

Best Chinese AI Models FAQ

What is the best Chinese AI model right now?

The answer box and first row above are the current decision receipt. They rebuild from the active public ranking lane, so use them instead of a copied winner sentence. When BenchAlign v5 is active, check the row’s evidence label and score interval before treating a narrow lead as decisive.

Which Chinese AI model is best for coding?

Use the coding leaderboard rather than this overall slice. Coding applies a different evidence mix, so its first row can differ from the broad leader shown here. A production choice should also account for repository language, tool use, latency, context needs, and the evidence available for that position.

Are the leading Chinese AI models open source?

Some leading rows are open weight and others are proprietary. Open weight means downloadable parameters are available under a model-specific license; it does not guarantee an OSI-approved license, unrestricted commercial use, reproducible training data, or inexpensive deployment. Read the license before choosing a self-hosted path.

Why is this page different from the Chinese LLM article?

This page owns the current ranking and regenerates with the data. The article is a dated analysis of why two scoring paths once produced different leaders. It remains useful as an audit trail, but it should not be read as a second live leaderboard or a competing recommendation page.

Last updated: July 20, 2026

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.