DashScope Model Lifecycle Reference 2026: Every Active, Deprecated, and Sunset Model in One Place

DashScope (Alibaba Cloud Model Studio / 百炼) now has five concurrent deprecation waves all converging on a single date: October 10, 2026 — just 79 days away. We have written individual migration guides for each wave, but until now there was no single reference page that ties everything together. This is that page.

OpenAI-compatible means a provider exposes a chat-completions endpoint whose request and response shape matches the OpenAI API contract closely enough that an unmodified OpenAI SDK call works against it after swapping three values: API key, base URL, and model name. The minimum surface in practice is POST /v1/chat/completions with messages, model, and an OpenAI-shaped streaming response.

DashScope Model Lifecycle Stages

Every model on DashScope goes through three stages:

1. Active — the model is available for API calls, receives updates, and is fully supported
2. Deprecated — the model has a published sunset date; QPM and TPM limits are gradually reduced; the API still works but you should migrate
3. Sunset (offline) — the model is removed; API calls return errors; no automatic redirect to replacements

Notification policy

Notifications are only sent to users with API call records in the previous 3 months. If you have dormant integrations, check the official deprecation page directly.

Active Text/Reasoning Models — July 2026

These are the current recommended models. If you are starting a new integration or migrating off a deprecated model, pick from this list.

Source: DashScope model pricing, retrieved 2026-07-23.

Deprecation Calendar: October 10, 2026 Mega-Sunset

All five deprecation waves converge on October 10, 2026 00:00:00 CST. Here is what is going away, organized by wave:

Wave 1 — Old/legacy models (notice 118434)

This wave covers the broadest set of models: legacy image generation, video generation, Qwen Plus/Flash/Long snapshots, Math, VL OCR, Omni, translation, speech synthesis/recognition, and third-party models.

Text/reasoning models affected:

Wave 2 — Mainline models (notice 118344)

Migration guide: Qwen3 to Qwen3.7 Upgrade Migration Guide

Wave 3 — Voice/speech mainline models (notice 118331)

Wave 4 — Historical mainline models (notice 118177)

This is the wave with the highest impact for text/reasoning API users:

Migration guides:

• October 2026 DashScope Massive Model Sunset Migration Guide
• Qwen Turbo/VL/Coder Plus Sunset Migration Guide

Wave 5 — Historical snapshots + third-party (notice 118345)

This wave removes dozens of specific snapshot versions and third-party models hosted on DashScope:

Key third-party models being removed:

Key Qwen snapshots being removed:

• All qwen3-max-* dated snapshots → qwen3.7-max
• All qwen3-vl-* model variants (8b, 30b, 32b, 235b) → qwen3.6-flash or qwen3.7-plus
• All qwen3-coder-* snapshots → qwen3.7-plus
• All qwen3-*b open-source snapshots → qwen3.6-flash or qwen3.7-plus
• All qwen2.5-* series snapshots → qwen3.6-flash or qwen3.7-plus

Migration guide: Migrating Off Bailian-Hosted DeepSeek, GLM, MiniMax

Already Sunset (for reference)

These models have already been removed and are no longer callable:

Decision Tree: Which Active Model to Pick

Need flagship reasoning?
  → qwen3.7-max (¥6/¥18 at promo, 1M context, thinking mode)

Need balanced quality + cost?
  → qwen3.7-plus (¥1.6/¥6.4 at promo, 1M context, thinking mode)

Need lowest cost?
  → qwen3.5-flash (¥0.2/¥2, 1M context, thinking mode)

Need fast inference?
  → qwen3.6-flash (¥1.2/¥7.2, 1M context)

Need multimodal (audio + vision)?
  → qwen3.5-omni-plus

Need coding-specific?
  → qwen3.7-plus (Coder Plus merges into Plus tier)

Need third-party model on DashScope?
  → kimi/kimi-k3 (2.8T params, 1M context)
  → glm-5.2 (Zhipu's latest)

TheRouter Integration: Handling Deprecated Model IDs

When DashScope sunsets a model, API calls to that model ID fail. TheRouter's model fallback routing lets you handle this without changing client code:

# Before: direct DashScope call with deprecated model
from openai import OpenAI

client = OpenAI(
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
    api_key="sk-xxx"
)
# This will fail after October 10, 2026:
response = client.chat.completions.create(
    model="qwen-turbo",
    messages=[{"role": "user", "content": "Hello"}]
)

# After: route through TheRouter with fallback
client = OpenAI(
    base_url="https://api.therouter.ai/v1",
    api_key="your-therouter-key"
)
# TheRouter routes to qwen3.7-plus when qwen-turbo is unavailable
response = client.chat.completions.create(
    model="qwen-turbo",
    messages=[{"role": "user", "content": "Hello"}]
)

TheRouter routes OpenAI-compatible requests through configured providers. You can set up fallback chains so that deprecated model IDs automatically route to their replacements, giving you time to update model references across your codebase.

Further Reading

Individual migration guides for each deprecation wave:

• October 2026 DashScope Massive Model Sunset Migration Guide
• September 2026 Qwen Model Sunset Migration Guide
• Qwen Turbo/VL/Coder Plus Sunset Migration Guide
• Migrating Off Bailian-Hosted DeepSeek, GLM, MiniMax
• Qwen3 to Qwen3.7 Upgrade Migration Guide
• DashScope Qwen3.7 Series Complete Guide

Provider and model pages:

• DashScope Provider Page
• Alibaba Cloud Bailian API Guide
• Chinese AI Model Deprecation Calendar Q3 2026

Sources:

• DashScope Model Deprecation Page — retrieved 2026-07-23
• DashScope Model Pricing — retrieved 2026-07-23
• DashScope Newly Released Models — retrieved 2026-07-23
• Official Notice 118434 — old/legacy models sunset
• Official Notice 118344 — mainline models sunset
• Official Notice 118331 — voice mainline models sunset
• Official Notice 118177 — historical mainline models sunset
• Official Notice 118345 — historical snapshots + third-party sunset