cocodot
Free trial
← Back to guides
Local card declined for overseas AI? cocodot: one card + one key
Try API freeOpen a cocodot card
Kimi K2.7Updated 2026-08

Kimi K2.7 Code API: Moonshot's Coding Model, Standard and High-Speed Tiers (2026)

K2.7 Code is Moonshot's programming-tuned model: 256K context, hardened instruction-following. Standard at ¥6.17/¥25.65 per M tokens, high-speed at double for interactive work. How it fits alongside K3 and Claude.

TL;DR: Kimi K2.7 Code is Moonshot AI's programming-tuned model: 256K context, markedly better instruction-following deep into long context, text/image/video input, thinking-mode output. cocodot runs both tiers at 95% of list: standard ¥6.17/¥25.65 (~$0.87/$3.6) per million tokens, high-speed ¥12.35/¥51.3 — same model on an accelerated serving stack, output much faster, built for interactive coding and agent loops where latency is real cost. Versus Moonshot's flagship K3 (1M context at ¥20/¥100): K2.7 is 3x cheaper and coding-tuned — K3 only earns its price when you need a whole repository in one window.

What K2.7 Code is good at

Two things: instruction-following that survives long context (models drift when you stuff a large project in — this one is tuned not to), and task completion rate — finishing the code rather than sketching it. Thinking-mode means it plans before writing, which suits agent-executor roles.

Is the high-speed tier worth 2x?

Depends where latency sits. Batch jobs and overnight pipelines: standard, half the price. Interactive coding in Cursor/Claude Code or multi-turn agent loops: high-speed — waiting time is developer cost, and doubled output speed usually beats doubled token price. Same model, switch per workload.

Setup and where it fits

OpenAI-compatible: base_url https://cocodot.co/api/ai/v1, model kimi-k2.7-code or kimi-k2.7-code-highspeed. A pragmatic stack: cheap tier for completions (Haiku/Sonnet 5/K2.7), Opus 5 for hard problems; K2.7 is the strongest Chinese-model alternative for code when you want a non-Anthropic lane — A/B it on your own workload with the same key.

Kimi coding lineup on cocodot (CNY per M tokens)

ModelIn / OutContextPosition
K2.7 Code standard¥6.17 / ¥25.65256Kcoding workhorse
K2.7 Code high-speed¥12.35 / ¥51.3256Ksame model, fast serving
K3 (reference)¥20 / ¥100 (list)1Mrepo-scale context

FAQ

K2.7 or K3?

For coding, K2.7 Code — 3x cheaper and tuned for it. K3 only when you need the 1M window for a whole repo or multi-day agent runs.

Image input?

Yes — text, image and video input; thinking-mode output.

Official model?

Official relay via a licensed cloud vendor; verify with probe.cocodot.co.

About cocodot

cocodot is a payment and AI access service for developers and cross-border teams in mainland China. It provides US-BIN virtual cards issued by a licensed institution — used to pay for overseas subscriptions and ad accounts — and an OpenAI-compatible AI API gateway for calling Claude, GPT and Gemini from within mainland China. Both share one wallet, funded by Alipay and accounted in USD. Card: $9.9 to open, 3% to load, $1 per active card per month; spending: $0.60 settlement fee on purchases under $20; a corresponding fee applies when the issuer charges one.

Service scope, pricing and limits →

Related reading

Claude API Returning 529 overloaded_error? How It Differs From 429 and What Actually Helpscontext_length_exceeded: Four Fixes, in the Order That Saves You Moneymodel not found: Five Causes and How to Tell Them Apart in Two MinutesCursor vs Claude Code vs Kiro vs Devin Desktop: Which AI Coding Tool Costs Less for How You Work (2026)

Calling Claude / GPT / Gemini?

Verify your email for $0.5 API credit — pay-as-you-go via Alipay, never above official price.

Try it free →

No card?

Open a cocodot card to pay for global AI subscriptions.

Learn about cocodot card →