Effort: max · Max
moonshotaiReleased Jul 16, 2026
Kimi-K3
2.8T total and 104B active parameters; combines Kimi Delta Attention with native image/video input under the Kimi K3 license.
Source description
Kimi K3 is Moonshot AI's 2.8T-parameter open-weight multimodal reasoning model. Built for complex coding, knowledge work, and long-horizon agentic workflows, it excels at navigating large repositories, calling tools, debugging, and iterating on images, logs, tests, and runtime feedback. Its architecture relies on KDA and Attention Residuals for computational efficiency.
Context window1M
Max output943.7K
Provider endpoints37
Benchmark results19
Measured performance
All leaderboards ↗Benchmark results
Effort: max · Max
GDPval-AA ↗2.1
1,532.59Effort: max · Max
AutomationBench-AA ↗AA Index 4.3.2 collection
58.27%Effort: max · Max
Effort: max · Max · Harness unreported
SciCode ↗AA Index 4.3.2 collection
59.49%AA · Under reviewEffort: max · Max
Humanity’s Last Exam ↗AA Index 4.3.2 collection
46.9%Effort: max · Max
GDP.pdf ↗AA Index 4.3.2 collection
22%Effort: max · Max
CritPt ↗AA Index 4.3.2 collection
23.43%AA · Under reviewEffort: max · Max
AA-Omniscience ↗AA Index 4.3.2 collection
19.7Effort: max · Max
AA-LCR ↗1.1
88.67%Effort: max · Max
LiveBench Overall ↗2.0 · 2026-10-14 (dataset 2026-09-02)
63.39%Effort: max · single run · equal mean of 7 suites
LiveBench Design ↗2.0 · 2026-10-14 (dataset 2026-09-02)
77.2%Effort: max · single run
LiveBench Research ↗2.0 · 2026-10-14 (dataset 2026-09-02)
79%Effort: max · single run
LiveBench Browser Use ↗2.0 · 2026-10-14 (dataset 2026-09-02)
16%Effort: max · single run
LiveBench Coding ↗2.0 · 2026-10-14 (dataset 2026-09-02)
59.3%Effort: max · single run
LiveBench Agentic ↗2.0 · 2026-10-14 (dataset 2026-09-02)
42.8%Effort: max · single run
LiveBench Data Analysis ↗2.0 · 2026-10-14 (dataset 2026-09-02)
78.7%Effort: max · single run
LiveBench Reasoning ↗2.0 · 2026-10-14 (dataset 2026-09-02)
90.7%Effort: max · single run
Provider measurements
1 configurationSpeed
≈10K input tokens1 requestmedian / 72h
Artificial Analysis · first-party providers · serving endpoint not reported
Kimi
Kimi K3 (Max)
Reasoning max
- Output rate
- 43.0872 tok/s
- Time to first chunk
- 3.66797 s
Last checked Source ↗
Methodology 2.2.0 & measurement details
First chunk includes request processing and network time, and may include reasoning.
- Prompt
- long · ≈10,000 input tokens
- Parallel requests
- 1
- Minimum answer
- 1,500 tokens
- Aggregation
- median over 72 hours
- Token counting
- o200k_base
- Benchmark client
- Google Cloud us-central1-a
- Serving endpoint
- Not reported
- Serving region
- Not reported
- Sample size
- Not reported
- Window end
- Not reported
Access & cost
38 quotesPrices across providers
Your token mix
USD · listed token ratesOpenRouter
Catalog quote- Input
- $0.304
- Output
- $11.936
- Cached input
- $0.21
together
via Hugging Face- Input
- $3.00
- Output
- $15.00
fireworks-ai
via Hugging Face- Input
- $3.00
- Output
- $15.00
featherless-ai
via Hugging Face- Input
- —
- Output
- —
baseten
via Hugging Face- Input
- $3.00
- Output
- $15.00
deepinfra
via Hugging Face- Input
- $2.85
- Output
- $14.25
- Input
- $2.85
- Output
- $14.25
Experiential
Standard list rate
- Input
- $3.00
- Output
- $15.00
- Cached input
- $0.23
Catalog list rate. Plan allowances and temporary discounts are separate.
Together AI
Serverless
- Input
- $3.00
- Output
- $15.00
- Cached input
- $0.30
Organization: Moonshot
Fireworks AI
Standard · service_tier=default
- Input
- $3.00
- Output
- $15.00
- Cached input
- $0.30
Fireworks AI
Standard · service_tier=priority
- Input
- $3.75
- Output
- $18.75
- Cached input
- $0.375
Fireworks AI
Fast · service_tier=default
- Input
- $4.50
- Output
- $22.50
- Cached input
- $0.45
- Input
- $0.25
- Output
- $8.00
- Cached input
- $0.18
- Input
- $0.28
- Output
- $16.00
- Cached input
- $0.27
- Input
- $0.30
- Output
- $14.90
- Cached input
- $0.238
- Input
- $0.304
- Output
- $11.936
- Cached input
- $0.21
- Input
- $0.36
- Output
- $11.90
- Cached input
- $0.359
- Input
- $0.504
- Output
- $8.10
- Cached input
- $0.18
- Input
- $0.84
- Output
- $13.50
- Cached input
- $0.30
- Input
- $1.20
- Output
- $14.00
- Cached input
- $1.20
- Input
- $1.53
- Output
- $12.75
- Cached input
- $0.204
- Input
- $1.65
- Output
- $8.25
- Cached input
- $0.165
Source discount 45%; rates as published
- Input
- $1.85
- Output
- $9.50
- Cached input
- $0.185
- Input
- $2.49
- Output
- $15.00
- Cached input
- $0.35
- Input
- $2.55
- Output
- $12.75
- Cached input
- $0.255
Source discount 15%; rates as published
- Input
- $2.55
- Output
- $12.95
- Cached input
- $0.255
- Input
- $2.85
- Output
- $14.25
- Cached input
- $0.285
- Input
- $3.00
- Output
- $15.00
- Cached input
- $0.30
Endpoint degraded (status -2)
- Input
- $3.00
- Output
- $15.00
- Cached input
- $0.30
Cache write $3.75/M tokens
- Input
- $3.00
- Output
- $15.00
- Cached input
- $0.30
Endpoint degraded (status -2)
- Input
- $3.00
- Output
- $15.00
- Cached input
- $0.30
- Input
- $3.00
- Output
- $15.00
- Cached input
- $0.30
- Input
- $3.00
- Output
- $15.00
- Cached input
- $0.30
- Input
- $3.00
- Output
- $15.00
- Cached input
- $0.30
- Input
- $3.00
- Output
- $15.00
- Cached input
- $0.30
- Input
- $3.45
- Output
- $17.25
- Cached input
- $0.345
- Input
- $4.50
- Output
- $22.50
- Cached input
- $0.45
- Input
- $4.50
- Output
- $22.50
- Cached input
- $0.45
Offers for this model
1 offerExperiential Labs · Discount
11.25% off · Experiential Pro model discounts
Paid Pro plan required, covering the first 2,000 credits of usage each month, then list price.
Checked Source ↗
Model specifications
- Developer
- moonshotai
- Modality
- Language
- Context
- 1,048,576 tokens
- Maximum output
- 943,718 tokens
- Weights
- Open weights
- Released
- Jul 16, 2026
- Provider listed
- Jun 13, 2026
- First discovered
- Oct 16, 2026
Exact identities
- openrouter
moonshotai/kimi-k3 - huggingface
moonshotai/Kimi-K3 - deepinfra
moonshotai/Kimi-K3 - experiential
kimi-k3 - artificial-analysis
f7d2fc3e-1f7b-405f-818c-07952a4af78f - livebench
kimi-k3
Sources
Updated Oct 18, 2026
More from moonshotai
View all ↗moonshotaiListed Jul 16
MoonshotAI: Kimi K3 (batch)
Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at...
1M contextImage inputVideo inputNo reported score
No reported score
$2.28in$11.40out/ 1M tokens · USDDeepInfra · via OpenRouter · 2 endpointsdeepinfra/bf16 · quantization=bf16
moonshotaiListed Jun 11
MoonshotAI: Kimi K2.7 Code
MoonshotAI: Kimi K2.7 Code is a coding-focused model in Moonshot AI's Kimi K2 family, built to complete end-to-end programming tasks reliably over long contexts. It uses a native multimodal…
262.1K contextImage inputTool callingNo reported score
No reported score
moonshotaiListed Apr 27
MoonshotAI: Kimi Latest
This model always redirects to the latest model in the Kimi family.
1M contextImage inputVideo inputNo reported score
No reported score
moonshotaiListed Apr 14
MoonshotAI: Kimi K2.6
Kimi K2.6 is Moonshot AI's next-generation multimodal model, designed for long-horizon coding, coding-driven UI/UX generation, and multi-agent orchestration. It handles complex end-to-end coding tasks…
262.1K contextImage inputTool callingNo reported score
No reported score