Kimi K3

Tool callingStructured outputReasoningImage inputVideo input

Hugging Face Launch post Website Docs

Multimodal Kimi model with 1M context and toggleable max-effort thinking for long-horizon agent work

Context: 1.0M
Released: 3mo ago
Parameters: 2.8T
Input / 1M tokens: $0.45
Output / 1M tokens: $16.50
Cached input / 1M tokens: $0.28

Sources:models.devOpenRouter

Benchmarks

DeepSWE is the primary coding benchmark where an exact model evaluation exists. Other sourced evaluations appear below; missing scores are not zero.

DeepSWE 1.1

Primary coding-agent benchmark · 113 tasks · Data generated 2026-09-22

Official leaderboard

68.5%

pass@1 · highest published configuration

mini-swe-agent · max reasoning

pass@4: 89.4%
Average cost / attempt: $4.65
Benchmark runs: 4

pass@1 is the success rate across scored attempts; pass@4 is the share of tasks solved in at least one of four attempts. Costs are reported by DeepSWE for each configuration. These results are not directly comparable to evaluations using different harnesses. Official source JSON.

Other published benchmarks

Evaluation	Reported score and source
ARC-AGI-2	55.0% Third-party evaluator · kimi-k3_high Reasoning effort: high External evaluation via Epoch AI Original-source terms apply
ARC-AGI-2	12.4% Third-party evaluator · kimi-k3_low Reasoning effort: low External evaluation via Epoch AI Original-source terms apply
ARC-AGI-2	60.4% Third-party evaluator · kimi-k3_max Reasoning effort: max External evaluation via Epoch AI Original-source terms apply
FrontierMath-Tier-4-v2-Private	39.0% Epoch AI · kimi-k3_max Reasoning effort: max Epoch AI evaluation
FrontierMath-Tiers-1-3-v2-Private	72.2% Epoch AI · kimi-k3_max Reasoning effort: max Epoch AI evaluation
GPQA diamond	91.9% Epoch AI · kimi-k3_high Reasoning effort: high Epoch AI evaluation
GPQA diamond	84.8% Epoch AI · kimi-k3_low Reasoning effort: low Epoch AI evaluation
GPQA diamond	93.1% Epoch AI · kimi-k3_max Reasoning effort: max Epoch AI evaluation
OTIS Mock AIME 2024-2025	93.3% Epoch AI · kimi-k3_high Reasoning effort: high Epoch AI evaluation
OTIS Mock AIME 2024-2025	68.9% Epoch AI · kimi-k3_low Reasoning effort: low Epoch AI evaluation
OTIS Mock AIME 2024-2025	97.2% Epoch AI · kimi-k3_max Reasoning effort: max Epoch AI evaluation
SimpleQA Verified	50.6% Epoch AI · kimi-k3_max Reasoning effort: max Epoch AI evaluation

Benchmark compilation: Epoch AI, Capabilities & benchmarking (CC BY attribution). Scores are converted to percentages and model names matched to the Vertical Tabs catalog. Individual evaluations are credited to their original sources. External results may carry separate licences and evaluation configurations; do not interpret scores from different harnesses as directly comparable.

Specifications

Developer: Moonshot AI
Model family: Kimi
Type: Language model
Predecessor: —
Successor: —
Released: July 16, 2026
Model ID: moonshotai/kimi-k3
Openness: Open weights
License: —
Parameters: 2.8T
Architecture: kimi-k3
Families: kimi-k3
Input modalities: text, image, video
Output modalities: text
Modality: text+image+video->text
Native context: 1.0M
Tokenizer: Other
Max output tokens: 944K
Hugging Face: moonshotai/Kimi-K3
Source freshness: Metadata synced Oct 8, 2026, 10:17 PM UTC

Open-weight builds

Ollama variants

Every published Ollama tag matched to this model, including parameter editions, instruction/base variants, and quantized builds.

1 builds

Variant	Parameters	Type	Quantization	Size	Context	Input
kimi-k3:cloud	2.8T	cloud	—	—	1M	text + image

Additional source evidence

Source model: moonshotai/kimi-k3
Source family: kimi-k3
Last updated: 2026-07-16
Model context: 1.0M
Model output limit: 131K
Open weights: Yes
File attachments: Yes
Reasoning: Yes
Tool calling: Yes
Structured output: Yes
Temperature control: No

Published benchmarks

Benchmark	Score	Evaluation details	Evidence
DeepSWE	67.5 · resolve rate	Kimi Code · max effort · 1.1 · 2026-07-16	Source
Terminal-Bench	88.3 · accuracy	Kimi Code · max effort · 2.1 · 2026-07-16	Source
FrontierSWE	81.2 · dominance score	Kimi Code · max effort · 2026-07-16	Source
Program Bench	77.8 · score	Kimi Code · max effort · 2026-07-16	Source
SWE Marathon	42 · resolve rate	Claude Code · max effort · 1.1 · 2026-07-16	Source
GDPval-AA	1668 · Elo	max effort · v2 · 2026-07-16	Source
AA-Briefcase	1548 · Elo	max effort · 2026-07-16	Source
AutomationBench	30.8 · success rate	600-task public subset · max effort · 2026-07-16	Source
JobBench	52.9 · score	max effort · 2026-07-16	Source
SpreadsheetBench	34.8 · score	Claude Code · max effort · 2 · 2026-07-16	Source
BrowseComp	91.2 · accuracy	max effort, context compaction · 2026-07-16	Source
CharXiv Reasoning	91.3 · accuracy	max effort, with tools · 2026-07-16	Source
ZeroBench	41 · pass@5	max effort, with tools · 2026-07-16	Source

Complete source record (JSON)

{
  "id": "moonshotai/kimi-k3",
  "name": "Kimi K3",
  "description": "Multimodal Kimi model with 1M context and toggleable max-effort thinking for long-horizon agent work",
  "family": "kimi-k3",
  "last_updated": "2026-07-16",
  "attachment": true,
  "reasoning": true,
  "tool_call": true,
  "structured_output": true,
  "temperature": false,
  "modalities": {
    "input": [
      "text",
      "image",
      "video"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 1048576,
    "output": 131072
  },
  "benchmarks": [
    {
      "name": "DeepSWE",
      "score": 67.5,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "resolve rate",
      "harness": "Kimi Code",
      "variant": "max effort",
      "version": "1.1",
      "date": "2026-07-16"
    },
    {
      "name": "Terminal-Bench",
      "score": 88.3,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "accuracy",
      "harness": "Kimi Code",
      "variant": "max effort",
      "version": "2.1",
      "date": "2026-07-16"
    },
    {
      "name": "FrontierSWE",
      "score": 81.2,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "dominance score",
      "harness": "Kimi Code",
      "variant": "max effort",
      "date": "2026-07-16"
    },
    {
      "name": "Program Bench",
      "score": 77.8,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "score",
      "harness": "Kimi Code",
      "variant": "max effort",
      "date": "2026-07-16"
    },
    {
      "name": "SWE Marathon",
      "score": 42,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "resolve rate",
      "harness": "Claude Code",
      "variant": "max effort",
      "version": "1.1",
      "date": "2026-07-16"
    },
    {
      "name": "GDPval-AA",
      "score": 1668,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "Elo",
      "variant": "max effort",
      "version": "v2",
      "date": "2026-07-16"
    },
    {
      "name": "AA-Briefcase",
      "score": 1548,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "Elo",
      "variant": "max effort",
      "date": "2026-07-16"
    },
    {
      "name": "AutomationBench",
      "score": 30.8,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "success rate",
      "variant": "max effort",
      "dataset": "600-task public subset",
      "date": "2026-07-16"
    },
    {
      "name": "JobBench",
      "score": 52.9,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "score",
      "variant": "max effort",
      "date": "2026-07-16"
    },
    {
      "name": "SpreadsheetBench",
      "score": 34.8,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "score",
      "harness": "Claude Code",
      "variant": "max effort",
      "version": "2",
      "date": "2026-07-16"
    },
    {
      "name": "BrowseComp",
      "score": 91.2,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "accuracy",
      "variant": "max effort, context compaction",
      "date": "2026-07-16"
    },
    {
      "name": "CharXiv Reasoning",
      "score": 91.3,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "accuracy",
      "variant": "max effort, with tools",
      "date": "2026-07-16"
    },
    {
      "name": "ZeroBench",
      "score": 41,
      "source": "https://www.kimi.com/blog/kimi-k3",
      "metric": "pass@5",
      "variant": "max effort, with tools",
      "date": "2026-07-16"
    }
  ],
  "release_date": "2026-07-16",
  "open_weights": true
}

Infrastructure

Provider availability

All providers

Provider	Endpoint	Context	Input /M tokens	Output /M tokens	Cached input /M tokens	Cache write /M tokens	Uptime 1d	Data handling
Wafer	wafer	1.0M	$0.29	$24.00	$0.28	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Wafer	waferBatch pricing	1.0M	$0.29	$24.00	$0.28	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Relace	relace/fp4	1.0M	$0.30	$14.00	$0.30	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Relace	relace/fp4Batch pricing	1.0M	$0.32	$14.00	$0.30	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Wafer	wafer/us	1.0M	$0.33	$14.90	$0.32	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Wafer	wafer/usBatch pricing	1.0M	$0.33	$14.90	$0.32	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
InferenceNet	inference-net/fp4	1.0M	$0.45	$15.00	$0.28	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
InferenceNet	inference-net/fp4Batch pricing	1.0M	$0.45	$15.00	$0.28	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Morph	morph/fp8	1.0M	$0.80	$14.90	$0.29	—	98.35%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Morph	morph/fp8Batch pricing	1.0M	$0.80	$14.90	$0.29	—	98.28%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Sail Research	sail-research/fp4	1.0M	$0.84	$13.50	$0.30	—	99.62%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Sail Research	sail-research/fp4Batch pricing	1.0M	$0.84	$13.50	$0.30	—	99.69%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
AkashML	akashml/fp4	1.0M	$1.20	$14.00	$1.20	—	97.19%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
AkashML	akashml/fp4Batch pricing	1.0M	$1.20	$14.00	$1.20	—	99.25%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Makora	makora/fp4	1.0M	$1.53	$12.75	$0.20	—	92.09%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Checked 2026-10-15
Makora	makora/fp4Batch pricing	1.0M	$1.53	$12.75	$0.20	—	98.69%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Checked 2026-10-15
Chutes	chutes/mxfp4	1.0M	$1.85	$9.50	$0.19	—	95.97%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Chutes	chutes/mxfp4Batch pricing	1.0M	$1.85	$9.50	$0.19	—	91.36%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Decart	decart/mxfp4	1.0M	$2.13	$10.65	$0.21	—	99.14%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Decart	decart/mxfp4Batch pricing	1.0M	$2.13	$10.65	$0.21	—	99.27%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
InferenceNet	inference-net/fast	1.0M	$2.49	$15.00	$0.35	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
InferenceNet	inference-net/fastBatch pricing	1.0M	$2.49	$15.00	$0.35	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
DigitalOcean	digitalocean	1.0M	$2.55	$12.95	$0.26	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
DigitalOcean	digitaloceanBatch pricing	1.0M	$2.55	$12.95	$0.26	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Phala	phala	1.0M	$2.55	$12.75	$0.26	—	99.79%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Phala	phalaBatch pricing	1.0M	$2.55	$12.75	$0.26	—	99.93%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Parasail	parasail/fp4	1.0M	$2.60	$13.00	$0.26	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Parasail	parasail/fp4Batch pricing	1.0M	$2.60	$13.00	$0.26	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
DeepInfra	deepinfra/mxfp4	1.0M	$2.85	$14.25	$0.29	—	99.93%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
DeepInfra	deepinfra/mxfp4Batch pricing	1.0M	$2.85	$14.25	$0.29	—	99.93%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Amazon Bedrock	amazon-bedrock/us-east-2	1.0M	$3.00	$15.00	$0.30	$3.75	95.91%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Amazon Bedrock	amazon-bedrock/us-east-2Batch pricing	1.0M	$3.00	$15.00	$0.30	$3.75	94.81%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
BaseTen	baseten/fp8	1.0M	$3.00	$15.00	$0.30	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
BaseTen	baseten/fp8Batch pricing	1.0M	$3.00	$15.00	$0.30	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Fireworks	fireworks	1.0M	$3.00	$15.00	$0.30	—	97.53%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Fireworks	fireworksBatch pricing	1.0M	$3.00	$15.00	$0.30	—	97.56%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Modal	modal/mxfp4	1.0M	$3.00	$15.00	$0.30	—	98.35%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Modal	modal/mxfp4Batch pricing	1.0M	$3.00	$15.00	$0.30	—	98.35%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Moonshot AI	moonshotai/mxfp4	1.0M	$3.00	$15.00	$0.30	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Moonshot AI	moonshotai/mxfp4Batch pricing	1.0M	$3.00	$15.00	$0.30	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Parasail	parasail/fast	1M	$3.00	$15.00	$0.30	—	97.67%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Parasail	parasail/fastBatch pricing	1M	$3.00	$15.00	$0.30	—	98.31%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Together	together	1.0M	$3.00	$15.00	$0.30	—	99.85%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Together	togetherBatch pricing	1.0M	$3.00	$15.00	$0.30	—	99.63%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Alibaba	alibaba	1.0M	$3.45	$17.25	$0.35	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Alibaba	alibabaBatch pricing	1.0M	$3.45	$17.25	$0.35	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Fireworks	fireworks/us	1.0M	$4.50	$22.50	$0.45	—	60.05%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Fireworks	fireworks/fast	1.0M	$4.50	$22.50	$0.45	—	99.93%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Fireworks	fireworks/usBatch pricing	1.0M	$4.50	$22.50	$0.45	—	45.67%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Fireworks	fireworks/fastBatch pricing	1.0M	$4.50	$22.50	$0.45	—	99.93%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15

Provider data:OpenRouter

Deployment graph

Can I run this?

Vertical Tabs only shows compatibility when it has structured evidence tying this exact model to hardware, runtime, quantization, memory use, and observed performance.

Browse hardware