Kimi K3
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 navigating large repositories, using tools, debugging, and iterating against images, logs, tests, and runtime feedback. Its architecture uses KDA and Attention Residuals for computational efficiency.
Model card on Hugging Face- Input / output price
- $1.85 / $9.50 per 1M
- Context
- 1M tokens
- Released
- Jul 16, 2026
Paste into Claude Code, Cursor, or Codex
Help me set up the Respan gateway: fetch and follow https:/
Providers
23 routes from 22 providers, cheapest first, OpenRouter last, priced per 1M tokens after gateway discounts.
| Provider | Input | Output |
|---|---|---|
Concentrate concentrate/kimi-k3 | $0.60/M | $15.00/M |
Wafer wafer/kimi-k3 | $0.60/M | $15.00/M |
Makora makora/kimi-k3 | $1.00/M | $18.00/M |
Chutes chutes/kimi-k3 | $1.85/M | $9.50/M |
Relace relace/kimi-k3 | $2.00/M | $14.00/M |
MorphLLM morphllm/morph-kimik3 | $2.50/M | $14.00/M |
IO.net ionet/moonshotai/Kimi-K3 | $2.7468/M | $16.50/M |
DeepInfra deepinfra/kimi-k3 | $2.85/M | $14.25/M |
RunAnywhere runanywhere/kimi-k3 | $2.925/M | $15.00/M |
Baseten baseten/kimi-k3 | $3.00/M | $15.00/M |
Bedrock bedrock/global.moonshotai.kimi-k3 | $3.00/M | $15.00/M |
EmpirioLabs empiriolabs/kimi-k3 | $3.00/M | $15.00/M |
Fireworks fireworks/kimi-k3 | $3.00/M | $15.00/M |
Inference.net inference_net/kimi-k3 | $3.00/M | $15.00/M |
Moonshot moonshot/kimi-k3 | $3.00/M | $15.00/M |
Nebius nebius/kimi-k3 | $3.00/M | $15.00/M |
Novita novita/kimi-k3 | $3.00/M | $15.00/M |
Parasail parasail/kimi-k3 | $3.00/M | $15.00/M |
Perplexity perplexity/kimi-k3 | $3.00/M | $15.00/M |
Tera tera/kimi-k3 | $3.00/M | $15.00/M |
TogetherAI togetherai/kimi-k3 | $3.00/M | $15.00/M |
Bedrock bedrock/us.moonshotai.kimi-k3 | $3.30/M | $16.50/M |
OpenRouter openrouter/kimi-k3 | $0.192/M | $16.00/M |
Is a provider down? AWS Bedrock status, Fireworks AI status, Kimi status, Perplexity status, OpenRouter status
Pricing
Averaged across 23 routes from 22 providers, in USD per 1M tokens after gateway discounts.
- Average input
- $2.46 per 1M
- Average output
- $14.95 per 1M
- Average cache read
- $0.45 per 1M
- Average cache write
- $2.91 per 1M
Cost by volume
Total cost from 0 to 1T tokens on the cheapest and most expensive provider.
Throughput and latency
30-day averages from requests through the Respan gateway. Under each figure, every model on the gateway by its average, this one in blue.
Output speed
Higher is better29tokens/sacross providers
Time to first token
Lower is better1.8sacross providers
Response latency
Lower is better15sacross providers
Benchmarks
The Respan Index plus 82 published scores from 14 sources, each linked to where it was published.
52.8/ 100
Respan Index#35 of 176
- 35Kimi K352.8
- 34Grok 4.753.0
- 35Qwen3.8-Max52.8
Published scores
| Benchmark | Score | Source | Published |
|---|---|---|---|
| Terminal-Bench 2.1github.com | 88.3% | github.com | Jul 27, 2026 |
| WeirdMLHåvard Tveit Ihle | 82.6% | Håvard Tveit Ihle | Oct 5, 2026 |
| LiveBench CodingLiveBench | 81.4% | LiveBench | Oct 5, 2026 |
| FrontierSWEgithub.com | 81.2% | github.com | Jul 27, 2026 |
| ProgramBenchgithub.com | 77.8% | github.com | Jul 27, 2026 |
| Kimi Code Bench 2.0github.com | 72.9% | github.com | Jul 27, 2026 |
| DeepSWE v1.1Datacurve | 68.5% | Datacurve | Oct 5, 2026 |
| DeepSWE v1.1 (Kimi Code)github.com | 67.5% | github.com | Jul 27, 2026 |
| SciCodegithub.com | 58.7% | github.com | Jul 27, 2026 |
| MirrorCodeEpoch AI | 54.4% | Epoch AI | Oct 31, 2026 |
| MLS-Bench-Litegithub.com | 48.3% | github.com | Jul 27, 2026 |
| SWE-Marathongithub.com | 42% | github.com | Jul 27, 2026 |
| PostTrainBenchgithub.com | 36.6% | github.com | Jul 27, 2026 |
| FrontierSWE V2Proximal Labs | 25.9% | Proximal Labs | Oct 5, 2026 |
| Terminal-Bench 4.0Vals AI | 17.2% | Vals AI | Oct 5, 2026 |
| Benchmark | Score | Source | Published |
|---|---|---|---|
| DeepSearchQA (F1)github.com | 95% | github.com | Jul 27, 2026 |
| MCPMark-Verifiedgithub.com | 94.5% | github.com | Jul 27, 2026 |
| BrowseComp (context compaction)github.com | 91.2% | github.com | Jul 27, 2026 |
| BrowseComp (1M context, no compaction)github.com | 90.4% | github.com | Jul 27, 2026 |
| OSWorld-Verifiedgithub.com | 84.8% | github.com | Jul 27, 2026 |
| MCP-Atlasgithub.com | 84.2% | github.com | Jul 27, 2026 |
| Toolathlon-Verifiedgithub.com | 76.5% | github.com | Jul 27, 2026 |
| Toolathlon-VerifiedHKUST | 76.5% | HKUST | Jul 16, 2026 |
| ResearchRubricsgithub.com | 76.2% | github.com | Jul 27, 2026 |
| OfficeQA Progithub.com | 63.3% | github.com | Jul 27, 2026 |
| LiveBench Agentic CodingLiveBench | 62.2% | LiveBench | Oct 5, 2026 |
| OSWorld 2.0github.com | 58.3% | github.com | Jul 27, 2026 |
| Finance Agent v2github.com | 54.4% | github.com | Jul 27, 2026 |
| JobBenchgithub.com | 54.3% | github.com | Jul 27, 2026 |
| APEX-AgentsMercor | 47.8% | Mercor | Oct 28, 2026 |
| APEX-Agentsgithub.com | 41% | github.com | Jul 27, 2026 |
| tau2-bench Banking KnowledgeSierra | 37.1% | Sierra | Aug 4, 2026 |
| tau3-Bankinggithub.com | 33.4% | github.com | Jul 27, 2026 |
| AutomationBenchgithub.com | 30.8% | github.com | Jul 27, 2026 |
| Agents' Last Examgithub.com | 28.3% | github.com | Jul 27, 2026 |
| Vending-Bench 2Andon Labs | 5165.04 · #35 of 68 | Andon Labs | Oct 5, 2026 |
| GDPval-AA v2github.com | 1686 Elo · #6 of 8 | github.com | Jul 27, 2026 |
| AA-Briefcasegithub.com | 1548 Elo | github.com | Jul 27, 2026 |
| Benchmark | Score | Source | Published |
|---|---|---|---|
| GPQA Diamondgithub.com | 93.5% | github.com | Jul 27, 2026 |
| GPQA DiamondEpoch AI | 93.1% | Epoch AI | Oct 5, 2026 |
| LiveBench ReasoningLiveBench | 90.7% | LiveBench | Oct 5, 2026 |
| LiveBench Data AnalysisLiveBench | 78.7% | LiveBench | Oct 5, 2026 |
| AA-LCRgithub.com | 74.7% | github.com | Jul 27, 2026 |
| SimpleBenchSimpleBench | 60.7% | SimpleBench | Jul 17, 2026 |
| Humanity's Last Exam (with tools)github.com | 56% | github.com | Jul 27, 2026 |
| Humanity's Last Exam (no tools)github.com | 43.5% | github.com | Jul 27, 2026 |
| Chess PuzzlesEpoch AI | 37% | Epoch AI | Oct 28, 2026 |
| Mystery Game PuzzlesEpoch AI | 26% | Epoch AI | Oct 5, 2026 |
| Benchmark | Score | Source | Published |
|---|---|---|---|
| OTIS Mock AIME 2024-2025Epoch AI | 97.2% | Epoch AI | Oct 5, 2026 |
| AIME 2026MathArena | 96.7% | MathArena | Sep 17, 2026 |
| LiveBench MathematicsLiveBench | 84.4% | LiveBench | Oct 5, 2026 |
| FrontierMath Tiers 1-3 v2Epoch AI | 72.2% | Epoch AI | Oct 5, 2026 |
| FrontierMath Tier 4 v2Epoch AI | 39% | Epoch AI | Oct 5, 2026 |
| Benchmark | Score | Source | Published |
|---|---|---|---|
| LiveBench LanguageLiveBench | 85.5% | LiveBench | Oct 5, 2026 |
| LiveBench Instruction FollowingLiveBench | 71.4% | LiveBench | Oct 5, 2026 |
| SimpleQA VerifiedEpoch AI | 50.6% | Epoch AI | Oct 5, 2026 |
| Benchmark | Score | Source | Published |
|---|---|---|---|
| MathVision (with Python)github.com | 97.8% | github.com | Jul 27, 2026 |
| MathVision (no tools)github.com | 94.3% | github.com | Jul 27, 2026 |
| CharXiv Reasoning (with Python)github.com | 91.3% | github.com | Jul 27, 2026 |
| Video-MME (with subtitles)github.com | 90% | github.com | Jul 27, 2026 |
| BabyVision (with Python)github.com | 85.7% | github.com | Jul 27, 2026 |
| CharXiv Reasoning (no tools)github.com | 84.8% | github.com | Jul 27, 2026 |
| MMMU-Pro (with Python)github.com | 83.4% | github.com | Jul 27, 2026 |
| MMVUgithub.com | 82.1% | github.com | Jul 27, 2026 |
| MMMU-Pro (no tools)github.com | 81.6% | github.com | Jul 27, 2026 |
| WorldVQA ForceAnswergithub.com | 51% | github.com | Jul 27, 2026 |
| ZeroBench pass@5 (with Python)github.com | 41% | github.com | Jul 27, 2026 |
| Furniture AssemblyEpoch AI | 34.2% | Epoch AI | Oct 5, 2026 |
| ZeroBench pass@5 (no tools)github.com | 23% | github.com | Jul 27, 2026 |
| Benchmark | Score | Source | Published |
|---|---|---|---|
| LMArena WebDevLMArena | 1658 Elo · #20 of 124 | LMArena | Nov 19, 2026 |
| LMArena EloLMArena | 1488 Elo · #19 of 191 | LMArena | Nov 19, 2026 |
| LMArena AgentLMArena | -0.0015 · #36 of 68 | LMArena | Nov 18, 2026 |
| Benchmark | Score | Source | Published |
|---|---|---|---|
| Harvey Lab-AAgithub.com | 94.6% | github.com | Jul 27, 2026 |
| OmniDocBenchgithub.com | 91.1% | github.com | Jul 27, 2026 |
| SAGE 1.1Vals AI | 86.8% | Vals AI | Oct 31, 2026 |
| CorpFin v2github.com | 71.6% | github.com | Jul 27, 2026 |
| SaaS-Benchgithub.com | 60.1% | github.com | Jul 27, 2026 |
| PerceptionBenchgithub.com | 58.5% | github.com | Jul 27, 2026 |
| SAGEVals AI | 52.8% | Vals AI | Oct 5, 2026 |
| Legal Research Benchgithub.com | 44.2% | github.com | Jul 27, 2026 |
| SpreadsheetBench 2github.com | 34.8% | github.com | Jul 27, 2026 |
| CritPtgithub.com | 23.4% | github.com | Jul 27, 2026 |
How Kimi compares: Claude vs Kimi, DeepSeek vs Kimi, Gemini vs Kimi
Compare Kimi K3: Kimi K3 vs Kimi K2.6, Kimi K3 vs GPT-6.1 Sol, Kimi K3 vs Kimi K3.1 Pro, Kimi K3 vs Kimi K2.7 Code, Kimi K3 vs Kimi K2.5, Kimi K3 vs Kimi K2
Apps
Use Kimi K3 in the tools you already run. Each guide points the app at the Respan gateway.
Setup guides for more apps are in the integration docs.
Kimi K3 FAQ
Short answers, built from the data on this page.
Kimi K3 is a model from Moonshot, released on Jul 16, 2026. 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 navigating large repositories, using tools, debugging, and iterating against images, logs, tests, and runtime feedback. Its architecture uses KDA and Attention Residuals for computational efficiency.
On the Respan gateway, input starts at $0.192 per 1M tokens on OpenRouter, and output starts at $9.50 per 1M tokens on Chutes, after gateway discounts. Averaged across its 23 routes from 22 providers, it costs $2.46 per 1M input tokens and $14.95 per 1M output tokens.
Kimi K3 is available through 23 routes from 22 providers: Concentrate, Wafer, Makora, Chutes, Relace, MorphLLM, and 16 more.
Kimi K3 accepts up to 1M tokens of context on its largest provider route.
Point any OpenAI-compatible SDK at https://gateway.respan.ai/v1 with your Respan API key and set the model to "moonshot/kimi-k3". To set it up with a coding agent, copy the agent setup prompt at the top of this page.
Kimi K3 scores 52.8 out of 100 on the Respan Index v1.1, ranking #35 of 176 models. By category: Coding 38.8, Agents 61.8, Reasoning 54.8, Knowledge & language 52.9, and Human preference 57.7. The Respan Index combines independent and lab-reported benchmark results into one score, using benchmark data collected by Respan.
Kimi K3 has 82 published benchmark scores, including Terminal-Bench 2.1 88.3%, DeepSearchQA (F1) 95%, and GPQA Diamond 93.5%. Each score links to where it was published; the data is collected by Respan.
Certified where it counts.
Built to meet the security and privacy standards that enterprise and healthcare teams require.
ISO 27001
The internationally recognized standard for information security management.
SOC 2
Secure, compliant management of your data across all of our systems.
GDPR
Operated under GDPR, the world's strictest standard for data privacy.
HIPAA
HIPAA compliant, with a BAA available for healthcare teams.