Organized by capability category, with open-source status, context length, and pricing noted. USD prices are list-price snapshots (2026-10); RMB prices are subject to vendors' official sites.
| Models | Vendor | Open Source | Context | Pricing (per million tokens) | Positioning |
|---|---|---|---|---|---|
| DeepSeek-V4.1-Flash-1017 | DeepSeek | Open source, MIT | 1M | ¥0.02–2 input / ¥4–8 output (peak/off-peak) | Current flagship workhorse, native vision understanding, switchable thinking mode |
| DeepSeek-V4-Pro-0813 | DeepSeek | Open source, MIT | 1M | ¥0.15–9 input / ¥13.5–27 output (peak/off-peak) | 1.6T total-parameter MoE flagship, deep thinking |
| Kimi K3.1 Pro | Moonshot AI | Open Source | 1M | $1.54 / $7.72 | Coding and long-horizon agent flagship, native multimodal |
| Kimi K3.1 Pro HighSpeed | Moonshot AI | Open Source | 1M | $3.09 / $15.44 | Same-generation intelligence with roughly 3× throughput in high-speed version |
| Kimi K3 | Moonshot AI | Open Source | 1M | $2.32 / $11.58 | 2.8T total parameters, Kimi Delta Attention architecture |
| Kimi K2.6 | Moonshot AI | API | 262K | $0.73 / $3.09 | General multimodal, long-horizon coding and self-correction |
| GLM-5.3 | Z.ai (Zhipu) | API | 1M | $1.08 / $3.40 | Flagship reasoning: software engineering and long-horizon Agents |
| GLM-5.3-Flash | Z.ai (Zhipu) | API | 1M | $0.12 / $0.39 | Native multimodal, high value for money; coding / Agents |
| GLM-5.3-FlashX | Z.ai (Zhipu) | API | 1M | $0.29 / $0.97 | Flash high-speed version for professional workflows |
| GLM-5.2 | Z.ai (Zhipu) | API | — | $1.08 / $3.40 | 744B MoE (~40B activated), released 2026-06-13 |
| Qwen3.7-Max | Alibaba | API | — | $1.27 / $3.82 | Closed-source flagship, long-horizon agents and complex reasoning |
| Qwen3.5–3.8 open-source family | Alibaba | Open source, Apache | — | Free for self-hosting / see Bailian | The world's most-downloaded open-source model family |
| Doubao Seed 2.1 Pro | ByteDance | API | — | $0.39 / $1.93 | Multimodal reasoning: Agents, coding, and visual understanding |
| Doubao-Seed-2.0 | ByteDance | API | 256K | See Volcano Ark | Same foundation as theDoubao app, #1 in consumer user base |
| Tencent Hunyuan Hy3 | Tencent | API | — | $0.10 / $0.41 | Foundation model of the WeChat ecosystem, highly cost-effective |
| Tencent Hunyuan Hy4 Preview | Tencent | Open Source | — | $0.64 / $1.93 | Hybrid-reasoning MoE: coding, Agents, long context |
| MiniMax M3.1 Flash | MiniMax | API | 1M | $0.12 / $0.46 | Multimodal reasoning + tool calling with adjustable reasoning effort |
| MiniMax M3 / M2.7 | MiniMax | API | 200K | $0.23 / $0.93 | Mainstay for reasoning and tool calling |
| ERNIE X1 / 4.5 Series | Baidu | 4.5 open-sourced | 128K | See Qianfan Platform | Fully in-house full-stack development, first open-sourced in 2025 |
| Spark X1 / 4.5 Ultra | iFLYTEK | API | — | See iFLYTEK Open Platform | A leading example of training on fully domestic compute |
| Step-3 | StepFun | Open Source | — | — | Multimodal MoE (321B), open-sourced in 2025 |
| SenseNova V6 | SenseTime | API | — | — | Flagship reasoning model with fused visual intelligence |
| LongCat-Flash | Meituan | Open Source | 128K | Free for self-hosting | A 560B MoE debut entry into open source |
| Ling Ling-1T | Ant Group | Open Source | — | Free for self-hosting | Ant's open-source trillion-parameter MoE |
| MiniCPM 4.0 | ModelBest | Open Source | — | Free (on-device) | On-device SOTA, runs locally on phones / PCs |
| Tiangong T1 | Kunlun Tech | API | — | — | Search-augmented reasoning with a multi-product portfolio |
| InternLM3 | Shanghai AI Laboratory | Open Source | — | Free for self-hosting | Flagship academic open-source model |
| Yi-Lightning | 01.AI | API | — | — | Lightweight, high-speed dialogue |
| Models | Vendor | Open Source | Context | Pricing | Positioning |
|---|---|---|---|---|---|
| DeepSeek-V4.1-Flash (Vision) | DeepSeek | Open source, MIT | 1M | Same-language version | Primary model with native image understanding, no model switching needed |
| GLM-5V-Turbo | Z.ai (Zhipu) | API | — | See bigmodel | Accelerated vision-language version |
| GLM-4.5V | Z.ai (Zhipu) | Open Source | — | Free for self-hosting | 106B-A12B open-source visual reasoning benchmark |
| Qwen2.5-VL / Qwen3-VL | Alibaba | Open Source | 128K | See Bailian | Open-source vision-language benchmark with screen-operation Agent |
| ERNIE 4.5-VL | Baidu | Open Source | 128K | Self-hosted / Qianfan | Natively cross-modal and open source |
| Kimi-VL-A3B | Moonshot AI | Open Source | — | Free | Small yet powerful lightweight visual reasoning |
| MiniCPM-V 4.5 | ModelBest | Open Source | — | Free | On-device multimodal SOTA |
| Doubao-1.5-vision-pro | ByteDance | API | — | See Volcano Ark | Image / video understanding |
| Models | Vendor | Open Source | Pricing | Positioning |
|---|---|---|---|---|
| Seedream 5.0 Lite | ByteDance | API | $0.027 / image | Built-in reasoning and web search, 3K resolution, multi-image editing |
| Seedream 4.5 | ByteDance | API | $0.031 / image | Text-to-image / image-to-image / sequential generation |
| Seedream 4.0 | ByteDance | API | $0.023 / image | Acclaimed Chinese text-to-image model |
| Qwen Image 2.1 Pro | Alibaba | API · 7B | $0.027 / image | Unified text-to-image and editing with native transparent RGBA output |
| Qwen Image 2.1 Turbo | Alibaba | API | $0.011 / image | 8-step accelerated checkpoints, high cost-effectiveness |
| GLM-Image | Z.ai (Zhipu) | API | See bigmodel | Z.ai (Zhipu) image generation line |
| HunyuanImage 3.0 | Tencent | Open Source | Free | 80B MoE, one of the largest open-source text-to-image models (2025) |
| CogView-4 | Z.ai (Zhipu) | Open Source | Free | Established open-source text-to-image line |
| Kolors | Kuaishou | Open source + API | Per image | High open-source visibility (2025) |
| ERNIE YiGe | Baidu | API | Per image | Established text-to-image model (2025) |
| Models | Vendor | Open Source | Pricing | Positioning |
|---|---|---|---|---|
| Seedance 2.5 | ByteDance | API | $8.26 / million output | Next-gen generation + editing + continuation, long-form multilingual |
| MiniMax H3 Max | MiniMax | API | $0.100 / clip | Text/image/reference-to-video, native audio, 2K |
| MiniMax H3 Max Turbo | MiniMax | API | $0.031 / clip | 480P/768P high-speed tier |
| MiniMax H3 Fast | MiniMax | API | $0.039 / clip | 480P fast tier |
| Wan 3.0 Video Prime | Alibaba | API | $0.216 / clip | Wan flagship tier, end-to-end speed boost |
| Wan 3.0 Video | Alibaba | API | $0.154 / clip | Text-to-video / first-last frame / reference generation, built-in audio |
| Vidu Q1 | Shengshu Tech | API | Custom | Cinematic-quality 1080p text/image-to-video |
| CogVideoX-3 | Z.ai (Zhipu) | Open source + API | See bigmodel | Powers the Qingying App |
| Kling | Kuaishou | API | Per second | Top-tier video generation globally; versions per official release |
| HunyuanVideo | Tencent | Open Source | Free | Lightweight open source, runs on consumer GPUs (2025) |
| SkyReels-V2 | Kunlun Tech | Open Source | Free | Unlimited-length long video generation (2025) |
| Berry 1.0 | CarrotHub | API | $0.15–0.22 / clip | Aggregated video generation based on Wan 3.0 |
| Models | Vendor | Open Source | Pricing | Positioning |
|---|---|---|---|---|
| Doubao Realtime Voice / Seed-TTS | ByteDance | API | Per character | Volcano Engine leads TTS market share |
| GLM-ASR-2512 | Z.ai (Zhipu) | API | See bigmodel | Speech recognition |
| MiniMax Speech 2.5 | MiniMax | API | Per character | Speech synthesis in 40+ languages |
| Qwen3-Omni / CosyVoice 2 | Alibaba | Open Source | Free | Open-source flagship for end-to-end omni-modal and TTS |
| Spark Voice Model | iFLYTEK | API | Per call / character | Deep focus on Chinese speech |
| Kimi-Audio | Moonshot AI | Open Source | Free | General audio understanding (2025) |
| Model / Product | Vendor | Open Source | Context | Positioning |
|---|---|---|---|---|
| Qwen3-Coder-480B | Alibaba | Open source, Apache | 256K–1M | Flagship open-source code model, agentic coding |
| GLM coding line / CodeGeeX | Z.ai (Zhipu) | Open Source | — | GLM-5 series coding flagship + veteran open-source assistant |
| Comate | Baidu | API | — | Enterprise-grade coding assistant |
| CodeBuddy | Tencent | API | — | Tencent Cloud AI Coding |
| Tongyi Lingma | Alibaba | API | — | Alibaba coding assistant |
| Models | Vendor | Open Source | Industry | Positioning |
|---|---|---|---|---|
| Pangu 5.5 | Huawei | Closed-source | Government / Mining / Meteorology | Benchmark industry LLM, full-stack Ascend |
| XuanYuan | Du Xiaoman | Open Source | Finance | Leading open-source financial LLM |
| Baichuan-M2-32B | Baichuan | Open Source | Healthcare | Open-source medical reasoning flagship |
| Spark Healthcare · Xiao Yi | iFLYTEK Health | Closed-source | Healthcare | Broad healthcare deployment coverage |
| Ziyue | NetEase Youdao | Closed-source | Education | Education-focused reasoning model |
| MathGPT | TAL Education | Closed-source | Math / Education | Strong at math problem-solving |
| MindVLA / World Foundation Model | Li Auto | In-house use | Autonomous driving | Representative VLA autonomous-driving LLM, no public API |
| Models | Vendor | Open Source | Pricing | Positioning |
|---|---|---|---|---|
| BGE-M3 + bge-reranker | BAAI | Open Source | Free | De facto standard for Chinese retrieval |
| text-embedding-v4 | Alibaba | API | ≈¥0.5 / million tokens | Vector service bundled with Alibaba Bailian |
| embedding-3 | Z.ai (Zhipu) | API | ≈¥0.5 / million tokens | LLM vendor's bundled vector offering |
| DeepSeek Search / Fetch | DeepSeek | API | ¥5 / 1,000 calls | Official web search and page-crawling tool |
Listing notes: USD pricing reflects the October 2026 catalog snapshot (TuneRouter official list prices); RMB pricing is based on each vendor's official website. Entries marked "2025" refer to previous-generation versions, provided for comparison. Suggestions for additions or revisions are welcome — please reach out.