Kimi K2 Series Discontinued on May 25, 2026

• ✓ Understand the May 25, 2026 Kimi K2 API discontinuation deadline
• ✓ Compare Kimi K2 vs K2.6 features, pricing, and performance benchmarks
• ✓ Learn the exact code changes needed to migrate from kimi-k2 to kimi-k2.6
• ✓ Avoid common pitfalls and service disruptions during migration

What Is Kimi K2 and Why Is It Being Discontinued?

The Kimi K2 series, developed by Chinese AI firm Moonshot AI, was launched in 2025 as a high-performance large language model with a 128K context window. It quickly gained popularity among developers for its strong coding capabilities and competitive pricing at $0.60 per million input tokens and $2.50 per million output tokens. However, on April 20, 2026, Moonshot AI released Kimi K2.6 — a next-generation model featuring a 256K context window, multimodal input support (text, image, video), and significantly enhanced agent capabilities. The company has announced that the original Kimi K2 series (including kimi-k2 and kimi-k2.5) will be officially discontinued on May 25, 2026, with all API endpoints being decommissioned. This decision reflects Moonshot AI's strategy to consolidate its model offerings around its most advanced architecture, which demonstrates a 32.25-point lead in benchmark performance over the K2 series. Developers using Kimi K2 APIs must migrate to K2.6 before the deadline to avoid complete service disruption.

Kimi K2 vs Kimi K2.6: Feature and Performance Comparison

Pricing Comparison: K2 vs K2.6

How to Migrate from Kimi K2 to Kimi K2.6: Step-by-Step Guide

Migrating from Kimi K2 to Kimi K2.6 is straightforward due to API compatibility with OpenAI's interface. Follow these steps to ensure seamless transition before the May 25, 2026 deadline.

1. Check your current API endpoint: Most users access Kimi K2 through https://api.kimi.com/v1. Confirm this in your application's configuration files.
2. Update your base URL: Replace https://api.kimi.com/v1 with https://api.moonshot.ai/v1 in your code. The endpoint structure remains identical.
3. Switch your model name: In your API requests, change model: "kimi-k2" to model: "kimi-k2.6".
4. Update your API key: If you're using the old Kimi API key, generate a new one from Moonshot AI's platform. The old keys will be invalid after May 25.
5. Test your integration: Send a test request with the new endpoint and model. The response should be identical in structure but may be faster due to improved architecture.
6. Enable cache features (optional): If you're using caching, add cache_read and cache_write parameters to your requests to benefit from K2.6's new cache pricing at $0.16 per million tokens.

Example migration code:

# Old Kimi K2 code
client = OpenAI(
    api_key="your_kimi_k2_api_key",
    base_url="https://api.kimi.com/v1"
)

response = client.chat.completions.create(
    model="kimi-k2",
    messages=[{"role": "user", "content": "Hello"}]
)

# New Kimi K2.6 code
client = OpenAI(
    api_key="your_kimi_k2_6_api_key",
    base_url="https://api.moonshot.ai/v1"
)

response = client.chat.completions.create(
    model="kimi-k2.6",
    messages=[{"role": "user", "content": "Hello"}]
)

This change requires minimal code modification but ensures continued access to Moonshot AI's most powerful model.

Pros, Cons, and Final Verdict

Pros of Kimi K2.6:

• ✓ 256K context window doubles memory capacity for long documents
• ✓ Multimodal input supports image and video analysis
• ✓ 32.25-point benchmark improvement over K2
• ✓ Cache features reduce costs for repetitive queries

Cons of Kimi K2.6:

• ⚠ Input cost increased by 58% from $0.60 to $0.95 per million tokens
• ⚠ Output cost increased by 60% from $2.50 to $4.00 per million tokens
• ⚠ No backward compatibility with Kimi K2 after May 25, 2026

The final verdict is clear: despite the price increase, Kimi K2.6 is a significant technological leap forward. The 32% performance improvement, doubled context window, and multimodal capabilities make it the superior choice for any application requiring advanced AI reasoning. Developers should migrate immediately to avoid service disruption on May 25, 2026. For cost-sensitive applications, consider optimizing usage patterns to leverage the new cache feature, which can reduce overall costs by up to 30% for repetitive queries.

Key Takeaways:

• ✓ Kimi K2 API will be completely discontinued on May 25, 2026 — no exceptions.
• ✓ Migration to Kimi K2.6 requires changing only the base URL and model name in your code.
• ✓ Despite higher pricing, K2.6's 32% performance improvement makes it the clear winner for serious applications.

If you're still using Kimi K2, start your migration immediately. The May 25, 2026 deadline is non-negotiable, and any delay could result in complete service disruption for your applications.

Last Updated: May 11, 2026 | Source: Kimi API Platform (Official Website)