AIHubMixAIHubMix
← Models

Kimi K2.5 vs Kimi K3

Output tokens cost $3.00 per million on Kimi K2.5 and $15.00 per million on Kimi K3. Input tokens cost $0.60 per million on Kimi K2.5 and $3.00 per million on Kimi K3. Cached input tokens are billed at $0.11 per million on Kimi K2.5 and $0.30 per million on Kimi K3. Context lengths are 256,000 tokens on Kimi K2.5 and 1,048,576 tokens on Kimi K3. Time to first token (TTFT) measured on AIHubMix is 1.1s on Kimi K2.5 and 0.9s on Kimi K3. Measured output throughput is 32.3 tok/s on Kimi K2.5 and 73.1 tok/s on Kimi K3. On the LMArena coding leaderboard Kimi K2.5 scores 1501 and Kimi K3 scores 1531.

Moonshot AIKimi K2.5Moonshot AIKimi K3
+ Add model · 2/4
Moonshot AI logo
Kimi K2.5
Moonshot AI · text, image, video → text

Kimi K2.5 is the smartest model of Kimi to date, achieving open-source state-of-the-art (SoTA) performance in Agent, coding, visual understanding, and a series of general intelligent tasks. At the same time, Kimi K2.5 is also the most versatile model of Kimi so far, with a native multimodal architecture design that supports both visual and text input, thinking and non-thinking modes, as well as dialogue and Agent tasks.

Input$0.60 /M
Output$3.00 /M
Details
Moonshot AI logo
Kimi K3
Moonshot AI · text, image, video → text

Kimi K3 is Kimi’s flagship model for long-horizon coding and end-to-end knowledge work, with a 1M-token context window and industry-leading intelligence.

Input$3.00 /M
Output$15.00 /M
Details

Pricing & Specifications

Prices are per million tokens. Time to First Token and throughput are rolling averages measured on AIHubMix.

Kimi K2.5
Kimi K3
Input /M
$0.60
$3.00
Output /M
$3.00
$15.00
Cache read /M
$0.11
$0.30
Context length
256,000
1,048,576
Max output
0
1,048,576
Time to First Token
1.1 s
0.9 s
Throughput
32.3 tok/s
73.1 tok/s
Modalities
textimagevideo
textimagevideo
Supported Parameters
thinkingfunction callingstructured outputs
thinkingfunction callingstructured outputs
API Formats
Released
-
-

Promotional prices show the discounted rate; see each model page for promotion windows.

Activity Past 30 Days

Daily traffic served through AIHubMix — how demand for each model is trending.

kimi-k2.5kimi-k3

Tokens / day

-

Requests / day

-

Performance Past 3 Days

Measured on real AIHubMix traffic, hourly buckets. Gaps mean no traffic in that hour.

kimi-k2.5kimi-k3

Throughput (tok/s)

-

TTFT (s)

-

Uptime (%)

-

LMArena Benchmarks

LMArena ratings by capability (Bradley-Terry, commonly called Elo). Higher is better.

Text
kimi-k2.5kimi-k3
14001440148015201560
Overall
14501486
Coding
15011531
Math
14701470
Hard prompts
14721505
Instruction following
14401479
Multi-turn
14521499
Creative writing
14221470
Longer query
14591502
Chinese
15021527
English
14611491

Source: LMArena (arena.ai) leaderboard, imported by AIHubMix. Models without published ratings are omitted per chart.

Cost calculator

Estimate your monthly bill for the same workload on each model.

Kimi K2.5
$81.00 /mo
Kimi K3
$405 /mo

Monthly = daily × 30. Discounted rates applied where a promotion is active.

FAQ

Which is cheaper: Kimi K2.5, Kimi K3?

Kimi K2.5: $3.00/M output tokens; Kimi K3: $15.00/M. Use the cost calculator above to estimate your own workload.

How do their coding arena scores compare?

Kimi K3: 1531; Kimi K2.5: 1501 (LMArena coding leaderboard).

Which responds faster?

Kimi K3: 0.9s time to first token measured on AIHubMix; see the live performance charts above for how each model behaves across the day.

How large is each context window?

Kimi K2.5 accepts 256,000 and Kimi K3 accepts 1,048,576 input tokens.

Which one generates tokens faster?

Kimi K3 at 73.1 tok/s and Kimi K2.5 at 32.3 tok/s, measured as output throughput on AIHubMix — a separate metric from time to first token.

What inputs and capabilities does each model support?

Kimi K2.5 accepts text, image and video input and supports thinking, function calling and structured outputs; Kimi K3 accepts text, image and video input and supports thinking, function calling and structured outputs.

Can I call Kimi K2.5 and Kimi K3 with the same API key?

Yes. AIHubMix serves every model on this page behind one OpenAI-compatible endpoint, so switching between them is a one-line change to the model field — no second account, key or SDK.

Popular comparisons

Related model match-ups readers also look at.

kimi-k3vsqwen3.8-max-previewclaude-opus-5vsgpt-5.6-lunaclaude-opus-5vsclaude-sonnet-5gemini-3.1-pro-previewvsgpt-5.5glm-5.2vskimi-k3gemini-3.6-flashvsqwen3.8-maxdeepseek-v4-flashvsglm-5.2claude-opus-5vsgrok-4.5
AIHubMix

© 2023 - 2026 AIHubMix, LLC

Product

Company

Developer

Connect