Moonshot AI language model

Kimi K3

Moonshot AI's most capable Kimi model: an open-weights 2.8T-parameter mixture-of-experts with native vision and a 1M-token context. it sits above Kimi 2.6 and Kimi 2.7 Code, which have a 256K context and cost less per token.

01

price

$3/M

input tokens

$15/M

output tokens

$0.3/M

cached input

per 1M tokens. octomind bills the listed provider price, with no markup, on every plan, including the free one (pay as you go from credits).

  • not included in any plan: every request bills from credits.
  • context caching is automatic: cache hits bill at $0.30 per 1M input tokens, a tenth of the input rate.
  • thinking is on by default at max effort, and reasoning tokens bill at the output rate.

02

what real work costs

workloadcost
one chat reply 3,000 tokens in, 500 out $0.02
summarise a 100-page report 50,000 tokens in, 1,500 out, one request $0.17
1,000 classification calls 500 tokens in, 20 out, per call $1.80
coding-agent session 2M tokens in across ~60 turns, 60,000 out, no cache $6.90
same session, cached 85% of input read from the prompt cache, the rest written to it $2.31

at the listed rate. token counts are typical sizes, not measurements.

03

specs

provider
Moonshot AI
context window
1M tokens — ≈ 1,500 pages of prose or 100,000 lines of code
input
text, images
output
text
structured output
yes — JSON schema
reasoning mode
yes
prompt caching
yes — cache hits at $0.3/M
weights
open — Kimi K3 License
released
July 2026
status
live
model id
kimi-k3

integration notes

  • temperature, top_p, n, presence_penalty and frequency_penalty are fixed; omit them from requests.
  • thinking cannot be disabled; reasoning_effort accepts low, high or max and defaults to max.
  • return the full assistant message, including reasoning_content and tool_calls, in multi-turn and tool-call flows, not just content.

04

benchmarks

as published by the maker; not measured by octomind.

benchmarkscore
GPQA Diamondsource93.5
HLE-Full (w/ tools)source56.0
Terminal-Bench 2.1source88.3
BrowseCompsource91.2
OSWorld-Verifiedsource84.8

05

when to pick it

pick it if

  • you want the model Moonshot calls its most capable
  • your prompts outgrow the 256K context of Kimi 2.6 and 2.7 Code
  • you resend long prompts, so cache hits bill at a tenth of input

06

where it sits on price

blended price $6/M at 3 input tokens per output token: 39 live language models cost less, 4 cost more.

  1. Gemini 3.1 Pro $4.5/M
  2. MiMo V2.6 Pro UltraSpeed $5.438/M
  3. Kimi K3 $6/M
  4. Claude Opus 5.5 $8/M
  5. Muse Wonder 1.0 $8/M

07

use Kimi K3 in octomind

one octohub login reaches every model in the catalog. pick this one per run, per session, or as your default.

run it
octomind login
octomind run -m octohub:kimi-k3
switch mid-session
/model octohub:kimi-k3
make it the default — config.toml
[model]
name = "octohub:kimi-k3"

08

compare Kimi K3

all model comparisons →

09

similar models

nearest live models by listed input price.

view all models →

10

Kimi K3 faq

how much does Kimi K3 cost?

Kimi K3 lists at $3/M input, $15/M output, $0.3/M cached input per 1M tokens. a typical coding-agent session (2M tokens in, 60,000 out) comes to $6.90, or $2.31 when 85% of the input is read from the prompt cache. octomind bills the listed provider price, with no markup, on every plan, including the free one (pay as you go from credits).

what is the context window of Kimi K3?

1M tokens, ≈ 1,500 pages of prose or 100,000 lines of code.

does Kimi K3 accept images?

yes. it takes text and images as input and returns text.

is Kimi K3 open source?

its weights are open under Kimi K3 License, so it can also be self-hosted.

when was Kimi K3 released?

July 2026, by Moonshot AI.

is Kimi K3 free?

no. Kimi K3 bills per token at $3/M input, $15/M output, $0.3/M cached input per 1M. octomind's free plan reaches it pay as you go from credits, so you pay only for the tokens you use, at the provider price. credits carry a 3% top-up fee.

how do i use Kimi K3 in octomind?

run octomind login once, then octomind run -m octohub:kimi-k3. inside a running session, /model octohub:kimi-k3 switches to it without losing context.

use Kimi K3 in octomind

choose a model when you start, then switch mid-session when the work changes.