model comparison

vs

Kimi 2.6 vs Kimi 2.7 Code

two Moonshot AI models, compared on listed price, real workload cost, context and capabilities.

the short answer

  • Kimi 2.6 and Kimi 2.7 Code list the same input and output prices.
  • Kimi 2.6 reads cached input for less ($0.16/M against Kimi 2.7 Code's $0.19/M), which decides the bill for agents that resend long context.

side-by-side specs

specKimi 2.6Kimi 2.7 Code
providerMoonshot AIMoonshot AI
input price$0.95/M$0.95/M
output price$4/M$4/M
cached input$0.16/M$0.19/M
blended price (3:1)$1.712/M$1.712/M
context window256k tokens256k tokens
inputtext, imagestext, images, video
structured outputyesyes
weightsopen (Modified MIT)open (Modified MIT)
releasedApril 2026—

same work, both models

workloadKimi 2.6Kimi 2.7 Code
one chat reply3,000 tokens in, 500 outsame$0.0049$0.0049
summarise a 100-page report50,000 tokens in, 1,500 out, one request$0.05$0.05
1,000 classification calls500 tokens in, 20 out, per call$0.56$0.56
coding-agent session2M tokens in across ~60 turns, 60,000 out, no cache$2.14$2.14
same session, cached85% of input read from the prompt cache, the rest written to itKimi 2.7 Code costs 1.1× as much as Kimi 2.6$0.80$0.85

at listed rates. token counts are typical sizes, not measurements.

Kimi 2.6 pricing notes

  • caching is automatic on repeated prompt prefixes; cache hits bill at $0.16 per 1M input tokens and there is no separate cache-write charge.

Kimi 2.7 Code pricing notes

  • cache hits bill at $0.19 per 1M input tokens; cache writes bill at the normal input rate, with no surcharge.

when to pick which

Kimi 2.6

Kimi K2.6 is Moonshot AI's open-weight multimodal model, a 1T-parameter mixture-of-experts with 32B active, with thinking and instant modes. it bills the same as Kimi 2.7 Code; Kimi K3 is the pricier 1M-context model.

pick it if

  • you want open weights (Modified MIT licence) with image input
  • you reuse long prompts: automatic cache hits bill at about a sixth of input
  • you run long agent or coding sessions, the work Moonshot pitches it for

skip it if

  • you need more than a 256K context window; Kimi K3 offers 1M

Kimi 2.7 Code

Moonshot AI's coding-focused Kimi model, built on Kimi K2.6 with the same architecture and the same price. an open-weights 1T-parameter MoE (32B active) that always reasons; Moonshot says it uses about 30% fewer thinking tokens than K2.6.

pick it if

  • you want Moonshot's coding model at the same rate as Kimi 2.6
  • you run long agentic coding sessions over a 256K context
  • you need open weights you could later self-host

skip it if

  • you want general writing or analysis; Moonshot points those at K2.6
  • you need replies without a thinking pass; reasoning cannot be disabled

faq

which is cheaper, Kimi 2.6 or Kimi 2.7 Code?

Kimi 2.6 and Kimi 2.7 Code list the same input and output prices. on a coding-agent session (2M tokens in, 60,000 out) Kimi 2.6 costs $2.14 and Kimi 2.7 Code $2.14.

can i switch between Kimi 2.6 and Kimi 2.7 Code mid-task?

yes. in an octomind session, /model octohub:kimi-2.6 and /model octohub:kimi-2.7-code switch models without losing the conversation, so you can draft on one and review on the other.

try either model

switch between models inside the same octomind session.

prices from the catalog of 2026-11-18