Chinese AI models
115 current AI models from 16 Chinese developers, grouped by what they do. 69 of them publish open weights. The highest-ranked on LMArena’s text leaderboard is Kimi K3, at rank 18.
Text and coding models (84)
| Model | Lab | Open weights | Price | LMArena rank |
|---|---|---|---|---|
| Kimi K3 | Moonshot AI | Yes | $3 / $15 per 1M tokens; $0.30 cached input | #18 |
| Qwen3.8 | Alibaba | Yes | $2 / $6 per 1M tokens (Qwen Cloud) | #23 |
| GLM-5.3 | Zhipu AI | Yes | $1.40 / $4.40 per 1M tokens ($0.26 cached input) | #26 |
| MiMo V2.6 Pro | Xiaomi | Yes | $0.435 / $0.87 per 1M tokens | #28 |
| DeepSeek V4.1-Flash | DeepSeek | Yes | $0.30 / $1.20 per 1M tokens (peak, cache miss); $0.15 / $0.60 off-peak | #34 |
| GLM-5.3-Flash | Zhipu AI | Yes | $0.15 input / $0.50 output per 1M tokens ($0.03 cached input) | #35 |
| GLM-5.2 | Zhipu AI | Yes | $1.40 / $4.40 per 1M tokens; $0.26 cached input | #37 |
| Qwen3.7 Max | Alibaba | — | $2.5 / $7.5 per 1M tokens; $0.5 implicit cached input (Alibaba Cloud Model Studio, International) | #41 |
| ERNIE 5.1 | Baidu | — | ¥0.004 / ¥0.018 per 1K tokens for input up to 32K, ¥0.006 / ¥0.022 for 32K-128K (Baidu Qianfan, CNY only; USD fields are a third-party conversion) | #50 |
| MiMo-V2.5-Pro | Xiaomi | Yes | $0.435 / $0.87 per 1M tokens | #51 |
| GLM-5.1 | Zhipu AI | Yes | $1.40 / $4.40 per 1M tokens | #53 |
| Qwen3.5 Max Preview | Alibaba | — | $1.2 / $6 per 1M tokens | #56 |
| Kimi K2.6 | Moonshot AI | Yes | $0.95 / $4.00 per 1M tokens | #59 |
| Qwen3.6 Max Preview | Alibaba | — | $1.03 / $6.16 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #60 |
| DeepSeek V4-Pro | DeepSeek | Yes | $1.32 / $3.96 per 1M tokens (peak, cache miss); $0.66 / $1.98 off-peak | #64 |
| GLM 5 | Zhipu AI | Yes | $1 / $3.2 per 1M tokens | #65 |
| Dola Seed 2 Pro | ByteDance | — | $0.50 / $3.00 per 1M tokens for prompts up to 128K (BytePlus ModelArk) | #68 |
| Qwen3.7 Plus | Alibaba | — | $0.40 / $1.60 per 1M tokens (Alibaba Cloud Model Studio list price; limited-time 20% discount: $0.32 / $1.28) | #70 |
| HY3 | Tencent | Yes | $0.13 / $0.53 per 1M tokens | #74 |
| Kimi K2.5 | Moonshot AI | Yes | $0.6 / $3 per 1M tokens | #78 |
| MiMo V2.6 Flash | Xiaomi | Yes | $0.14 / $0.28 per 1M tokens | #81 |
| MiMo V2 Pro | Xiaomi | — | $1 / $3 per 1M tokens (before retirement on 30 June 2026) | #83 |
| Step 5 Preview | StepFun | — | $1 / $2.70 per 1M tokens | #84 |
| Qwen3.6 Plus | Alibaba | — | $0.5 / $3 per 1M tokens (Alibaba Cloud Model Studio, Singapore, 0-256K input tier) | #89 |
| Qwen3.5 397B A17b | Alibaba | Yes | $0.6 / $3.6 per 1M tokens (Alibaba Cloud Model Studio, Singapore, 0-256K input tier) | #92 |
| GLM 4.7 | Zhipu AI | Yes | $0.60 / $2.20 per 1M tokens | #95 |
| MiniMax-M3 | MiniMax | Yes | $0.30 / $1.20 per 1M tokens (standard, permanent 50% off list price) | #97 |
| DeepSeek V4 Flash High Preview | DeepSeek | Yes | $0.44 / $1.32 per 1M tokens (third-party listing via LMArena, not a DeepSeek price; model retired 10 September 2026) | #98 |
| Qwen3.8 27B | Alibaba | Yes | $0.50 / $3 per 1M tokens | #102 |
| DeepSeek V4-Flash | DeepSeek | Yes | $0.15 / $0.60 per 1M tokens off-peak; $0.30 / $1.20 peak — figures are DeepSeek-V4.1-Flash, which now serves this model | #105 |
| Qwen3 Max Preview | Alibaba | — | $0.78 / $3.9 per 1M tokens | #108 |
| MiMo V2.5 | Xiaomi | Yes | $0.14 / $0.28 per 1M tokens | #109 |
| GLM 5V Turbo | Zhipu AI | — | $1.2 / $4 per 1M tokens | #110 |
| MiMo V2 Omni | Xiaomi | — | $0.4 / $2 per 1M tokens (before retirement on 30 June 2026) | #115 |
| Kimi K2 Thinking Turbo | Moonshot AI | Yes | $1.15 / $8 per 1M tokens | #117 |
| GLM 4.6 | Zhipu AI | Yes | $0.60 / $2.20 per 1M tokens | #127 |
| Qwen3 Max | Alibaba | — | $0.78 / $3.9 per 1M tokens | #128 |
| Qwen3 235B A22b Instruct 2507 | Alibaba | Yes | $0.26 / $1.06 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #130 |
| Kimi K2 0905 Preview | Moonshot AI | Yes | $0.6 / $2.5 per 1M tokens | #133 |
| Kimi K2 0711 Preview | Moonshot AI | Yes | $0.6 / $2.5 per 1M tokens | #136 |
| Qwen3.5 122B A10b | Alibaba | Yes | $0.4 / $3.2 per 1M tokens (Alibaba Cloud Model Studio, Singapore, 0-256K input tier) | #138 |
| MiniMax M2.7 | MiniMax | Yes | $0.3 / $1.2 per 1M tokens | #141 |
| Qwen3 VL 235B A22b Instruct | Alibaba | Yes | $0.21 / $1.90 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #147 |
| GLM 4.5 | Zhipu AI | Yes | $0.6 / $2.2 per 1M tokens | #151 |
| Hunyuan HY3 Preview | Tencent | Yes | $0.29 / $1.17 per 1M tokens | #154 |
| Qwen3.5 27B | Alibaba | Yes | $0.3 / $2.4 per 1M tokens (Alibaba Cloud Model Studio, Singapore, 0-256K input tier) | #155 |
| Qwen3 235B A22b No Thinking | Alibaba | Yes | $0.46 / $1.82 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #161 |
| Qwen3 235B A22b Thinking 2507 | Alibaba | Yes | $0.23 / $2.30 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #165 |
| Qwen3 Next 80B A3b Instruct | Alibaba | Yes | $0.09 / $1.10 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #166 |
| DeepSeek R1 | DeepSeek | Yes | Open weights — not served by DeepSeek's API | #167 |
| Qwen3.5 Flash | Alibaba | — | $0.10 / $0.40 per 1M tokens (Alibaba Cloud Model Studio, International) | #168 |
| Qwen3 VL 235B A22b Thinking | Alibaba | Yes | $0.40 / $4 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #170 |
| Qwen3.5 35B A3b | Alibaba | Yes | $0.25 / $2 per 1M tokens (Alibaba Cloud Model Studio, Singapore, 0-256K input tier) | #172 |
| Step 3.5 Flash | StepFun | Yes | $0.1 / $0.3 per 1M tokens | #174 |
| MiniMax M2.5 | MiniMax | Yes | $0.30 / $1.20 per 1M tokens | #178 |
| Qwen3 Coder 480B A35b Instruct | Alibaba | Yes | $0.40 / $1.60 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #182 |
| MiMo V2 Flash | Xiaomi | Yes | $0.1 / $0.3 per 1M tokens (before retirement on 30 June 2026) | #186 |
| MiniMax M2.1 Preview | MiniMax | Yes | $0.3 / $1.2 per 1M tokens | #187 |
| Qwen3 30B A3b Instruct 2507 | Alibaba | Yes | $0.05 / $0.19 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #190 |
| GLM 4.6V | Zhipu AI | Yes | $0.3 / $0.9 per 1M tokens | #193 |
| GLM 4.5 Air | Zhipu AI | Yes | $0.20 / $1.10 per 1M tokens | #198 |
| Qwen3 Next 80B A3b Thinking | Alibaba | Yes | $0.15 / $1.20 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #200 |
| MiniMax M1 | MiniMax | Yes | $0.55 / $2.2 per 1M tokens | #205 |
| GLM 4.5V | Zhipu AI | Yes | $0.6 / $1.8 per 1M tokens | #217 |
| Step 3 | StepFun | Yes | $0.57 / $1.42 per 1M tokens | #220 |
| Qwen3 32B | Alibaba | Yes | $0.08 / $0.28 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #225 |
| Qwen Plus 0125 | Alibaba | — | $0.4 / $1.2 per 1M tokens | #228 |
| MiniMax M2 | MiniMax | Yes | $0.30 / $1.20 per 1M tokens | #229 |
| Ling Flash 2 | Ant Group | Yes | Open weights — free to self-host | #230 |
| QwQ 32B | Alibaba | Yes | $0.50 / $1 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #242 |
| Qwen3 30B A3b | Alibaba | Yes | $0.13 / $0.52 per 1M tokens (third-party listing; not on Alibaba Cloud Model Studio's pricing page) | #252 |
| Ring Flash 2 | Ant Group | Yes | Open weights — free to self-host | #258 |
| GLM 4 Plus | Zhipu AI | — | $0.44 / $1.76 per 1M tokens | #264 |
| Qwen Max 0919 | Alibaba | Yes | $1.6 / $6.4 per 1M tokens | #265 |
| Qwen2.5 72B Instruct | Alibaba | Yes | $1.20 / $1.20 per 1M tokens (earlier listing; Alibaba marks the Qwen2.5 series deprecated and no longer shows a price) | #287 |
| Kimi K3.1 Pro | Moonshot AI | — | $2 / $10 per 1M tokens; $0.20 cached input | — |
| Spark X2.6 | iFLYTEK | Yes | Not published in a form readable here — see xinghuo.xfyun.cn | — |
| Kimi K2.8 Preview | Moonshot AI | — | Kimi Code membership — not on the Kimi API price list | — |
| Hy4 preview | Tencent | Yes | $0.834 / $2.501 per 1M tokens; $0.042 cached input (Tencent Cloud, Singapore) | — |
| Qwen3.8-Flash-Next | Alibaba | Yes | Weights free under qwen-community-1.0; Qwen3.8-Flash API $0.15 / $0.47 per 1M tokens (Alibaba Cloud Model Studio International, qwen3.8-flash, a separate API product from the open weights) | — |
| Ling 3.0 Flash | Ant Group | — | $0.0168 / $0.0504 per 1M tokens (limited-time 72% discount; list $0.06 / $0.18) | — |
| Kimi K2.7 Code | Moonshot AI | Yes | $0.95 / $4 per 1M tokens; $0.19 cached input | — |
| Step 3.7 Flash | StepFun | Yes | $0.20 / $1.15 per 1M tokens; $0.04 cached input | — |
| LongCat-2.0 | Meituan | — | $0.75 / $2.95 per 1M tokens; $0.015 cached input (limited-time $0.30 / $1.20) | — |
Video generation models (20)
| Model | Lab | Open weights | Price | LMArena rank |
|---|---|---|---|---|
| Seedance 2.0 | ByteDance | — | $0.07 per second (480p), $0.15 (720p), $0.37 (1080p), $0.78 (4K) for a 5-second 16:9 clip (BytePlus ModelArk) | #5 |
| Seedance 2.5 | ByteDance | — | $0.103 per second (480p), $0.231 (720p), $0.569 (1080p) for a 5-second 16:9 clip (BytePlus ModelArk) | #7 |
| MiniMax H3 | MiniMax | Yes | $0.08 per second (768P), $0.13 per second (2K) on the MiniMax API; open weights under the MiniMax H3 Community License | #8 |
| Wan 2.7 T2V | Alibaba | — | $0.10 per second (720p), $0.15 per second (1080p) (Alibaba Cloud Model Studio, International) | #20 |
| Wan 2.6 T2V | Alibaba | — | $0.10 per second (720p), $0.15 per second (1080p) (Alibaba Cloud Model Studio, International) | #21 |
| Seedance 1.5 Pro | ByteDance | — | $2.40 per 1M tokens with audio, $1.20 without; half price offline (BytePlus ModelArk) | #22 |
| Kling 2.6 Pro | Kuaishou | — | Credits-based — see klingai.com | #29 |
| Hailuo 2.3 | MiniMax | — | $0.28 per 6-second 768P video, $0.56 per 10-second 768P, $0.49 per 6-second 1080P (MiniMax API) | #32 |
| Kling O1 Pro | Kuaishou | — | Credits-based — see klingai.com | #33 |
| Hailuo 02 Pro | MiniMax | — | $0.10 / $0.15 per 6s / 10s 512P video, $0.28 / $0.56 at 768P, $0.49 per 6s 1080P (MiniMax API, Hailuo-02, legacy) | #34 |
| Hunyuan Video 1.5 | Tencent | Yes | Open weights under the Tencent Hunyuan Community License — free to self-host | #38 |
| Vidu Q4 Preview | Shengshu Technology | — | 78 credits per second (4K) — see platform.vidu.com/docs/overview/pricing | — |
| Wan 3.0 | Alibaba | — | $0.05 per second (480p), $0.10 (720p), $0.20 (1080p) list price; limited-time 30% off (Alibaba Cloud Model Studio, International) | — |
| HappyHorse 1.1 | Alibaba | — | $0.07 per second (480p), $0.14 (720p), $0.18 (1080p) (Alibaba Cloud Model Studio, International) | — |
| Kling 3.0 Turbo | Kuaishou | — | $0.112 (720p) / $0.14 (1080p) per second with native audio (Kling API) | — |
| HappyHorse 1.0 | Alibaba | — | $0.14 per second (720p), $0.24 per second (1080p) (Alibaba Cloud Model Studio, International) | — |
| Kling Video 3.0 | Kuaishou | — | Free Basic tier; Standard $6.99/mo (660 credits); Pro $25.99/mo (3,000 credits); Premier $64.99/mo (8,000 credits); Ultra $127.99/mo (26,000 credits) - first-subscription monthly rates, annual billing cheaper | — |
| Vidu Q3 Pro | Shengshu Technology | — | Credits-based — see platform.vidu.com/docs/overview/pricing | — |
| Kling 2.5 Turbo | Kuaishou | — | 25 credits per 5-second 1080p video (Kling AI) | — |
| Wan 2.7 I2V | Alibaba | — | $0.10 per second (720p), $0.15 per second (1080p) (Alibaba Cloud Model Studio, International) | — |
Image generation models (10)
| Model | Lab | Open weights | Price | LMArena rank |
|---|---|---|---|---|
| Qwen Image 3.0 Pro | Alibaba | — | $0.04 per output image plus $0.003 per input image at 1K (Alibaba Cloud Model Studio, International) | #12 |
| Seedream 5.0 Pro | ByteDance | — | $0.045 per image up to 1.5K, $0.09 above (BytePlus ModelArk) | #14 |
| Qwen Image 2.0 Pro | Alibaba | — | $0.075 per image (Alibaba Cloud Model Studio, International) | #21 |
| Hunyuan Image 3.0 | Tencent | Yes | Open weights under the Tencent Hunyuan Community License — free to self-host | #33 |
| Seedream 4.5 | ByteDance | — | $0.04 per output image; input images free (BytePlus ModelArk) | #35 |
| Seedream 5.0 Lite | ByteDance | — | $0.035 per output image; input images free (BytePlus ModelArk) | #39 |
| Wan 2.6 T2I | Alibaba | — | $0.03 per image (Alibaba Cloud Model Studio, International) | #41 |
| HiDream O1 Image | HiDream | Yes | Open weights — free to self-host | #51 |
| Wan 2.7 Image Pro | Alibaba | — | $0.075 per image (Alibaba Cloud Model Studio, International) | #55 |
| GLM-Image | Zhipu AI | Yes | $0.015 per image (Z.ai API) | — |
Audio models (1)
| Model | Lab | Open weights | Price | LMArena rank |
|---|---|---|---|---|
| MiniMax Speech-02 | MiniMax | — | $100 per 1M characters (speech-02-hd), $60 per 1M characters (speech-02-turbo) (MiniMax API) | #6 |
Frequently asked questions
Which Chinese companies make AI models?+
On this site: Alibaba (40), Zhipu AI (14), MiniMax (10), Moonshot AI (9), Xiaomi (7), ByteDance (7), DeepSeek (5), Tencent (5), Kuaishou (5), StepFun (4), Ant Group (3), Shengshu Technology (2), Baidu (1), HiDream (1), Meituan (1), iFLYTEK (1).
Which Chinese AI model ranks highest on LMArena?+
Of the Chinese text models here, Kimi K3 from Moonshot AI has the highest rank on LMArena’s text leaderboard, at 18. The full table is at aiblogly.com/benchmarks.
Are Chinese AI models open source?+
69 of the 115 Chinese models listed here publish open weights, among them Kimi K3, Qwen3.8, GLM-5.3, MiMo V2.6 Pro, DeepSeek V4.1-Flash, GLM-5.3-Flash. Each licence is on the model’s page.