Kimi
Historical multimodal Kimi with 256K context, retired on the Kimi platform August 31, 2026. Migration reference for K3.
At a glance
API model ID: kimi-k2.5
Capabilities & boundaries
Tool support requires the appropriate API integration; a supported tool is not automatically active in every chat.
The historical model supported thinking and non-thinking modes. These are archival capabilities, not instructions for a currently available endpoint. K3 always uses thinking with its own effort controls; review the replacement's documentation before migrating.
Put it to work
Original migration and workflow examples from EZ Ai Assist. Use these with an available model and the appropriate tools; the retired system cannot run new requests.
Workflow 01
Find migration-sensitive assumptions before changing IDs.
Audit this older K2.5 integration using the code and API logs provided. Identify assumptions about model availability, thinking mode, input formats, tool history and token budgets. Separate confirmed incompatibilities from items needing a test. Produce a migration checklist without executing requests or treating the retired endpoint as available.
Workflow 02
Preserve useful behavior without relying on a retired model.
Turn these representative K2.5 conversations and expected results into a replacement-model evaluation set. Define pass criteria, allowed variation, factuality checks and escalation rules. Remove secrets from examples. Explain how to compare a new model with the recorded baseline without inventing results or requiring new calls to the retired model.
Workflow 03
Avoid carrying product promises into an API migration.
Review this workflow description and label each capability as model behavior, application orchestration, external tool access or human approval. Cite the evidence supporting each label. Flag any claim of agent swarms or automated actions that lacks an implemented integration. End with the smallest set of components needed for a safe replacement workflow.
Developer reference
Historical Kimi API information, not EZ Ai Assist subscription prices.
View EZ Ai Assist plansK2.5 is retired on the Kimi platform, so no current purchasable API rate is presented. The archived pricing page is a historical reference, not a live quote. Use K3's current table when estimating a migration.
Common questions
No. The current model list and changelog identify retirement on August 31, 2026. The changelog says requests return 404 model not found. A retained website page does not override that lifecycle notice.
Kimi points to K3. Test your actual prompts and tool loop because K3 has different reasoning controls, completion settings and cache-write billing. This guide links to the K3 page for that comparison.
It preserves an existing public URL and supports audits of older workflows. The historical label distinguishes a useful reference from an endorsement for a new deployment.
The archived provider pricing reference lists 262,144 tokens. This guide corrects the older catalog shorthand while keeping that capacity explicitly historical.
Do not equate Kimi's product-level Agent Swarm with a raw model call. Orchestrating multiple agents requires an application system; it is not an automatic feature of the retained EZ Ai Assist catalog entry.
An archived rate does not establish that you can buy access now. The page retains the source for historical comparison but does not advertise a retired endpoint as a current offering.
This guide describes the provider API, not subscription entitlements or a promise that every setting is exposed in the app. Check your workspace for model access and supported inputs, and use the EZ Ai Assist pricing page for plan details.
Check the source
Specifications and API prices checked on . Example prompts and workflow advice are editorial guidance from EZ Ai Assist.
Same provider
Kimi
Multimodal Kimi with 1M context, always-on thinking and adjustable reasoning effort for coding and knowledge work.
Kimi
Coding-focused Kimi with 256K context, visual inputs and always-on thinking for multi-step software work.
Kimi
Faster-serving K2.7 Code endpoint with the same model behavior, 256K context and separately priced API usage.
Kimi
Multimodal Kimi with 256K context and optional thinking for coding, document analysis and visual review.