Kimi K3

API ID kimi-k3

kimi-k3 is a Moonshot AI model, released July 2026. Standard price: $3.00 input and $15.00 output per 1M tokens.

Status: Active
Released: July 2026
Shutdown: none announced
Replacement: —

No deprecation or shutdown date has been announced as of October 15, 2026.

Benchmarks

Scores as each source published them

Comparing with Kimi K2.6, the previous model in this line. A score is compared only with the same benchmark from the same source, at the same setting where both have it. Full comparison: prices, limits, features

Arena

Arena	Kimi K3	Kimi K2.6	Change	Rank	Date	Source
Text By category Math 1500 rank 17 vs 1479 (+21, within range) at max effort (vs no setting) Creative writing 1460 rank 29 vs 1433 (+27) at max effort (vs no setting) Instruction following 1488 rank 15 vs 1456 (+32) at max effort (vs no setting) Hard prompts 1517 rank 11 vs 1486 (+31) at max effort (vs no setting) Long prompts 1503 rank 11 vs 1478 (+25) at max effort (vs no setting) Multi-turn 1499 rank 11 vs 1461 (+38) at max effort (vs no setting) Expert 1533 rank 10 vs 1503 (+30) at max effort (vs no setting)	1488 ±5	1461 ±4	+27	17 of 415 32,036 votes	Oct 13, 2026	source
Text: Coding	1539 ±7	1516 ±7	+23	10 of 410 8,136 votes	Oct 13, 2026	source
WebDev	1655 ±6	1509 ±7	+146	14 of 142 18,644 votes	Oct 13, 2026	source
Agent	0.019 ±0.007	not rated		25 of 54 134,561 sessions	Oct 14, 2026	source

Rated as kimi-k3-max (max effort); Kimi K2.6 as kimi-k2.6. Some were rated at different settings; ratings on one arena share one scale, so the difference still reads directly. Text and WebDev ratings come from people voting between two anonymous answers; Text is style-controlled, as on arena.ai. Agent is Arena's separate score for real agent sessions, on its own scale. Not rated on Vision and Document, where Kimi K2.6 is. Data: Arena (arena.ai), CC BY 4.0.

Epoch AI

Benchmark	Kimi K3	Kimi K2.6	Change	Setting	Run	Source
FrontierMath Tier 4	39.0%	25.6%	+13.4 pts	max effort (vs no setting)	Jul 17, 2026	source
FrontierMath Tiers 1-3	72.2%	57.2%	+15.0 pts	max effort (vs no setting)	Jul 17, 2026	source
GPQA Diamond Other settings 91.9% vs not run high effort 84.8% vs not run low effort	93.1%	90.8%	+2.3 pts	max effort (vs no setting)	Jul 16, 2026	source
OTIS Mock AIME 2024-2025 Other settings 93.3% vs not run high effort 68.9% vs not run low effort	97.2%	96.1%	+1.1 pts	max effort (vs no setting)	Jul 16, 2026	source
SimpleQA Verified	50.6%	34.9%	+15.7 pts	max effort (vs no setting)	Aug 27, 2026	source

Run independently by Epoch AI. Data: Epoch AI, Capabilities & benchmarking (epoch.ai/benchmarks), CC BY.

On cloud platforms

All platform dates

Platform	Platform ID	State	Shutdown	Time left	Replacement	Source
Amazon Bedrock	moonshotai.kimi-k3	Active	none announced		—	source

Read daily from each platform's own pages. The shutdown date and countdown on this page use Moonshot AI's own date.

Price history

USD per 1M tokens, standard tier

We started recording this model's prices on Oct 5, 2026. The history grows with each daily check.

Show as a table

From	Type	Price
Oct 5, 2026	Input	$3.00
Oct 5, 2026	Output	$15.00

Prices

USD per 1M tokens, every type and band, by tier

Type	Price $/1M	In effect
Standard
input	$3.00
cached input	$0.30
cache write (5 min)	$3.00
cache write (1 hour)	$6.00
output	$15.00

Changelog

Changes we detected on Moonshot AI's pages

No changes recorded yet.