AI Model Watch
Deprecations / Migrate

Migrate off Kimi Thinking (Preview)

Retired

Kimi Thinking (Preview) has been retired and shut down on 11 Nov 2025. Moonshot's official replacement is Kimi K2.6. The cheapest comparable model we track is Llama 3.1 8B Instruct (Meta) at $0.022 per 1M tokens blended.

Last verified: 13 Aug 2026 · sourced from official provider documentation

StatusRetiredMoonshot
Shut down11 Nov 2025plan ahead
Official replacementKimi K2.6$1.71 /M blendedCheapest alternativeLlama 3.1 8B Instruct$0.022 /M blended
What changes

Kimi Thinking (Preview) → Kimi K2.6

Moving to Kimi K2.6 means a larger context window (131K → 262K).

SpecKimi Thinking (Preview)Kimi K2.6
StatusRetiredGA
Blended price /1M—$1.71
Input / output /1M—$0.95 / $4
Context window131K262K
Max output——
Modalitytexttext, image, video
Knowledge cutoff——

API string was kimi-thinking-preview → switch to kimi-k2.6. See Kimi K2.6's full spec sheet →

Cheapest alternatives

Budget-friendly replacements for Kimi Thinking (Preview)

Generally-available models that match Kimi Thinking (Preview)'s capabilities (and at least half its 131K context), ranked by blended price. Every number is computed from official catalog prices.

ModelBlended /1MInput / outputContext
Llama 3.1 8B Instruct
Meta
$0.022$0.02 / $0.03128K
Amazon Nova Micro
Amazon
$0.061$0.035 / $0.14128K
Command R7B
Cohere
$0.066$0.037 / $0.15128K
FAQ

Migrating off Kimi Thinking (Preview)

What replaces Kimi Thinking (Preview)?

Moonshot's recommended migration for Kimi Thinking (Preview) is Kimi K2.6. The cheapest comparable model in our catalog is Llama 3.1 8B Instruct (Meta) at $0.022/1M blended.

Is Kimi Thinking (Preview) still available?

No — Kimi Thinking (Preview) has been retired (shut down 11 Nov 2025) and is no longer available via Moonshot's API.

What is the cheapest alternative to Kimi Thinking (Preview)?

Llama 3.1 8B Instruct (Meta) is the cheapest generally-available alternative we track that matches Kimi Thinking (Preview)'s capabilities, at $0.02 per 1M input and $0.03 per 1M output tokens ($0.022/1M blended).

Browse the full deprecation watch → or see Kimi Thinking (Preview)'s full spec sheet →