Nothing about glm-5.3-flashx is obvious the first time you see it. Chinese model ids follow no shared standard: each vendor encodes its own product organisation into the string, and Western-facing gateways pass those strings through unchanged. This is a decoder for the catalog we serve.
Almost every id is built from the same four pieces, in this order:
| Layer | Example | Question it answers |
|---|---|---|
| Vendor / family | glm, kimi, doubao, qwen | Who made it? |
| Product line | seedream, seedance, hy-3d, hy-mt | What is it for? |
| Generation | v4, 5.3, k2.7, m2.5 | How new is it? |
| Variant suffix | -flash, -pro, -turbo, -code | Which trade-off inside that generation? |
The confusion comes from vendors ordering these differently: ByteDance puts the product line inside the name (doubao-seedance-2.5), Alibaba mixes an internal generation number into the prefix (qwen3.7-max), and Tencent sometimes drops the vendor entirely (hy3 instead of hunyuan-3).
| Suffix | Means | Price signal within a family | Example |
|---|---|---|---|
-flash | Cheap, high-throughput tier | Lowest paid tier | glm-5.3-flash at $0.08/1M in |
-flashx | Branded sub-variant of flash | Between flash and full | glm-5.3-flashx at $0.12/1M in |
-lite | Smallest tier | Lowest | doubao-seed-2.0-lite at $0.075/1M in |
-pro | Quality tier | Higher than flash | deepseek-v4-pro at $0.57/1M in |
-max / -plus | Flagship tier | Highest in family | qwen3.8-max, qwen-mt-plus |
-turbo | Speed-optimised variant | Family-dependent | glm-5-turbo, doubao-seed-2.1-turbo |
-highspeed | Throughput-optimised | Often more expensive | kimi-k2.7-code-highspeed at $2.07/1M in |
-code | Tuned for code | Tag, not tier | doubao-seed-2.0-code |
-vision / -vl | Accepts images | Tag, not tier | deepseek-v4-flash-vision, qwen3-vl-plus |
-thinking | Exposes reasoning steps | Tag, not tier | glm-4.1v-thinking-flash |
-omni | Multiple input/output modalities | Tag, not tier | qwen3-omni-flash |
-preview | Pre-release snapshot | No guarantee | hy4-preview, step-5-preview |
-mt | Machine translation line | Cheapest per token | hy-mt2-lite at $0.043/1M in |
-asr / -tts | Speech recognition / synthesis | Billed differently | hy-asr-3.0-preview, qwen3-tts-flash |
-t2i | Text-to-image | Per image | wan2.2-t2i-flash |
-3d | 3D generation line | Per generation | hy-3d-motion, tripo-3d-3.1 |
-embedding | Embedding model | Per 1M, no output price | qwen3.7-text-embedding |
| Vendor | Chat / LLM | Image | Video | Other lines |
|---|---|---|---|---|
| DeepSeek | deepseek-v4-flash, -pro, v4.1-flash | — | — | -vision variant |
| Zhipu AI | glm-5.x (+ flash/flashx/turbo) | cogview-3-flash | cogvideox-flash | glm-4.6v vision |
| Alibaba Cloud | qwen3.7-max, qwen-plus, qwen3-coder-flash | wan2.x-t2i-* | — | qwen-mt-plus translation, qwen3-tts-flash, OCR, embeddings |
| Moonshot | kimi-k3, kimi-k2.x (+ -code) | — | — | -highspeed variant |
| ByteDance | doubao-seed-2.x (lite/pro/code/turbo) | seedream-* | seedance-* | — |
| Tencent Hunyuan | hy3, hy4-preview, hunyuan-* | hy-image-v3 | hy-video-v1.5 | hy-mt2-* translation, hy-3d-*, ASR |
| MiniMax | minimax-m2.5/m2.7/m3, minimax-text-01 | — | minimax-video-h3 | minimax-speech-*, minimax-music-* |
| Xiaomi | mimo-v2.5-pro, mimo-v2.6-flash | — | — | — |
| Kuaishou | — | — | kling-video-v3 | -omni, -turbo variants |
| Others | step-5-preview | wand-vega-image-*, vidu-image-q2 | pixverse-video-*, vidu-video-q3 | tripo-3d-*, Kinfra embeddings |
Flash means cheap, not free. GLM-5.3-Flash is $0.08 per 1M input; GLM-5.3 is $1.25. A handful of flash models (GLM-4.7-Flash, GLM-4.6V-Flash, GLM-4.1V-Thinking-Flash) happen to be free to call on our free tier — that is a plan decision, not a naming rule.
Inside ByteDance, Pro costs more than Lite. Across vendors the ordering collapses: doubao-seed-2.0-pro is $0.40 per 1M input, while kimi-k3 is $2.90 — a "Pro" that costs a seventh of another vendor's flagship. Rank across the whole catalog, not within one family.
ByteDance's image line ships as both doubao-seedream-5.0 and seedream-image-v5.0-pro. Same line, two conventions: one for the Doubao console, one for the standalone product. If your code hard-codes one and the other is retired, requests fail — prefer ids you can resolve from /v1/models at runtime.
step-5-preview, hy4-preview, kimi-k2.8-preview: preview snapshots can change weights or disappear. Fine for evaluation, risky as a production dependency.
GET /v1/models (or the catalog page).Flash marks the cheap, high-throughput tier of a family — not literally free. GLM-5.3-Flash is $0.08 per 1M input tokens while GLM-5.3 is $1.25. Some flash models (GLM-4.7-Flash, GLM-4.6V-Flash) are genuinely free to call on NovAI's free tier.
They are ByteDance product lines, not model generations: Doubao-Seed-* is the chat/LLM line, Seedream is image generation, Seedance is video generation.
Aliases. doubao-seedream-5.0 and seedream-image-v5.0-pro refer to the same ByteDance image line under different naming conventions; providers expose both, and gateways may list both to match what users already have in their code.
Within one family, almost always. Across families it says nothing. Doubao-Seed-2.0-Pro is $0.40 per 1M input, cheaper than Kimi K3 at $2.90, because families price differently. Compare across the catalog, not just within a family.
Free tier for the first 10 calls a day. No card. Prices come from the same normalized feed this guide cites.
Get an API keyModel ids, suffixes and prices reflect the live catalog on 2026-10-22. Verify any specific price at /china-ai-leaderboard.json; list the models your key can call at /v1/models.