// head_to_head

Kimi K2.6 vs Kimi K2.7 Code

Published LiveBench scores across all seven categories, live list pricing, context windows, and the measured cost of a point of capability — for both models, side by side.

OpenRouter + LiveBenchAll comparisonsFull leaderboard

Kimi K2.6 wins outright — it scores higher and costs less.

Kimi K2.6 leads by 2.1 points overall while listing 34% cheaper per blended million tokens. There is no trade-off to reason about here; the only reason to pick the other is a constraint this table does not show, like an existing contract, a region, or a provider you are already on. The two cost measures disagree here, which is worth knowing: Kimi K2.6 has the lower sticker price, but Kimi K2.7 Code earns each point of capability for less — $0.0545 against $0.0918 — because per-token rates do not predict how many tokens a model actually spends on a task.

moonshotai

Kimi K2.6

Blended / 1M
$1.01
Context
262K
Released
Apr 20, 2026
Overall score
70.5
reasoningtool callingimage inputprompt caching

moonshotai

Kimi K2.7 Code

Blended / 1M
$1.35
Context
262K
Released
Jun 12, 2026
Overall score
68.4
reasoningtool callingimage inputprompt caching

Specs and pricing

MetricKimi K2.6Kimi K2.7 Code
LiveBench overall

Mean of the seven LiveBench category scores, 0–100. Higher is better.

70.5win68.4
Cost per point

Measured benchmark spend divided by overall score — dollars per point of capability.

$0.0918$0.0545win
Blended price / 1M

3:1 input:output mix, the usual shape of production traffic.

$1.01win$1.35
Input price / 1M$0.560win$0.670
Output price / 1M$2.36win$3.40
Cached input / 1M

Price of an input token served from the prompt cache, where the provider publishes one.

$0.094win$0.170
Context window262K262K
Max output tokens262K262K

Benchmarks by category

An overall score averages away the thing you probably care about. These are the seven LiveBench categories scored separately — Kimi K2.6 on top, Kimi K2.7 Code below, both out of 100.

Agentic coding
46.9
45.7
Coding
78.6
74.0
Reasoning
79.4
82.8
Mathematics
84.3
79.6
Data analysis
65.1
62.7
Language
75.1
77.9
Instruction following
64.4
56.3

What each one costs to run

Per-token prices are hard to feel. These are monthly list costs for both models across five workload shapes, using each provider's published cached-input rate where there is one.

WorkloadKimi K2.6Kimi K2.7 Code
Support chatbot

1.2K in / 400 out × 200K requests

$289.76/mo$396.80/mo
RAG assistant

8K in / 600 out × 100K requests

$403.56/mo$540.00/mo
Coding agent

40K in / 4K out × 20K requests

$376.18/mo$528.00/mo
Document extraction

20K in / 1.5K out × 50K requests

$714.19/mo$900.00/mo
Bulk classification

500 in / 20 out × 5M requests

$1404.20/mo$1765.00/mo
Run these two through the cost calculator

Which should you pick?

You are running this at volume

Kimi K2.7 Code

Lowest measured cost per point of capability at $0.0545 per point — the gap compounds with every request.

Quality matters more than the bill

Kimi K2.6

Highest overall LiveBench score of the two at 70.5.

The workload is coding or agentic work

Kimi K2.6

Leads on agentic coding — 46.9 against 45.7.

Kimi K2.6 vs Kimi K2.7 Code FAQ

Which is better, Kimi K2.6 or Kimi K2.7 Code?

Kimi K2.6 wins outright — it scores higher and costs less. Kimi K2.6 leads by 2.1 points overall while listing 34% cheaper per blended million tokens. There is no trade-off to reason about here; the only reason to pick the other is a constraint this table does not show, like an existing contract, a region, or a provider you are already on. The two cost measures disagree here, which is worth knowing: Kimi K2.6 has the lower sticker price, but Kimi K2.7 Code earns each point of capability for less — $0.0545 against $0.0918 — because per-token rates do not predict how many tokens a model actually spends on a task.

Is Kimi K2.6 cheaper than Kimi K2.7 Code?

Kimi K2.6 is cheaper. On a 3:1 input:output blend, Kimi K2.6 lists at $1.01 per million tokens and Kimi K2.7 Code at $1.35 — Kimi K2.6 is 34% cheaper. Input and output are priced separately — Kimi K2.6 charges $0.560 in and $2.36 out, Kimi K2.7 Code charges $0.670 and $3.40 — so the model that looks cheaper flips depending on how output-heavy your workload is.

Kimi K2.6 vs Kimi K2.7 Code: which scores higher on benchmarks?

Kimi K2.6 scores 70.5 and Kimi K2.7 Code scores 68.4 overall on LiveBench, the mean of its seven categories. That is a 2.1-point lead for Kimi K2.6. Category scores differ from the overall figure — a model can lead on reasoning and trail on coding, which the per-category table above breaks out.

Which gives better value for money, Kimi K2.6 or Kimi K2.7 Code?

Kimi K2.7 Code. Cost per point divides the measured dollars LiveBench spent running the benchmark by the score it earned, so it captures something token pricing misses: a reasoning model can emit many times more tokens than its per-token rate suggests. Kimi K2.6 works out at $0.0918 per point and Kimi K2.7 Code at $0.0545.

Does Kimi K2.6 or Kimi K2.7 Code have a bigger context window?

They are effectively the same — 262K for Kimi K2.6 and 262K for Kimi K2.7 Code.

Do Kimi K2.6 and Kimi K2.7 Code support prompt caching?

Both publish a cached-input rate: $0.094 per million for Kimi K2.6 and $0.170 for Kimi K2.7 Code, against full input rates of $0.560 and $0.670. On a workload with a long stable prefix — a system prompt, a tool schema, a retrieved corpus — that changes the economics more than the headline price does.

Related comparisons

How these numbers are produced

  • Price — provider list price from OpenRouter, refreshed every 15 minutes. “Blended” is a 3:1 input:output mix.
  • ScoresLiveBench release 2026-06-25, using their own category map. Each model shows its strongest published run. A blank means “not evaluated”, never “bad”.
  • Cost per point — the measured dollars LiveBench spent on the run, divided by the score it earned.
  • “Win” — awarded only past a threshold: one full point on a benchmark score, 10% on a price, 25% on a context window. Anything tighter reports as a tie, because effort settings alone move a LiveBench score by more than that.

Published benchmarks rank models on someone else's tasks. Before committing, see LLM & agent evaluation for building an eval on your own.