Moonshot
Kimi K3
Launched 2026-07-16 · Listed on OpenRouter 2026-07-16
Prices updated . Benchmarks updated .
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...
You send / 1M
$0.192
tokens in
It replies / 1M
$16.00
tokens out
Context
1M
tokens it can read at once
Typical request
$0.0082
1,000 words in, 400 out
Benchmarks
| Benchmark | Score | Kind | Date | Source |
|---|---|---|---|---|
| Epoch Capabilities Index | 158.3 | Independent | 2026-07-16 | Epoch AI |
| GPQA DiamondMax effort | 93.1% | Independent | 2026-07-16 | Epoch AI |
| OTIS Mock AIMEMax effort | 97.2% | Independent | 2026-07-16 | Epoch AI |
| SimpleQA VerifiedMax effort | 50.6% | Independent | 2026-08-27 | Epoch AI |
| ARC-AGI-2Max effort | 60.4% | Independent | 2026-07-16 | ARC Prize, republished by Epoch AI |
| GDP.pdfMax effort | 19.0% | Independent | 2026-07-16 | Surge AI GDP.pdf, republished by Epoch AI |
| CritPtMax effort | 23.4% | Independent | 2026-07-16 | CritPt, republished by Epoch AI |
| Arena Elo (text)Max effort | 1,476 | Independent | 2026-10-28 | LMArena |
| Arena Elo (webdev)Max effort | 1,656 | Independent | 2026-10-28 | LMArena |
Rankings on this site use the Epoch Capabilities Index, an independent composite. Lab-reported rows are the lab's own figures and are kept separate.
Data privacy
No training or retention: Yes. At least one provider serves it with zero data retention.
| Provider | You send / 1M | It replies / 1M | Uptime | Privacy policy |
|---|---|---|---|---|
| Wafer · mxfp4 | $0.276 | $9.60 | 100% | Read |
| Relace · fp4 | $0.182 | $11.01 | 100% | Read |
| DigitalOcean | $2.55 | $12.95 | 100% | Read |
| Io Net · fp4 | $2.10 | $14.00 | 100% | Read |
| DeepInfra · mxfp4 | $2.85 | $14.25 | 100% | Read |
| Nebius · fp4 | $3.00 | $15.00 | 100% | Read |
| River | $3.00 | $15.00 | 100% | Read |
| InferenceNet · fp4 | $0.234 | $15.00 | 100% | Read |
| Amazon Bedrock | $3.00 | $15.00 | 100% | Read |
| InferenceNet · fp4 | $0.234 | $15.00 | 100% | Read |
| Parasail · fp4 | $3.00 | $15.00 | 100% | Read |
| Together | $3.00 | $15.00 | 99% | Read |
| Fireworks | $3.00 | $15.00 | 98% | Read |
| BaseTen · fp8 | $3.00 | $15.00 | 100% | Read |
| Moonshot AI | $3.00 | $15.00 | 100% | Read |
| Wafer · mxfp4 | $0.153 | $16.00 | 100% | Read |
| Wafer · mxfp4 | $0.192 | $16.00 | 100% | Read |
| InferenceNet · fp4 | $0.450 | $18.00 | 100% | Read |
| InferenceNet · fp4 | $0.450 | $18.00 | 100% | Read |
| Makora · fp4 | $1.00 | $18.00 | 98% | n/a |
| Parasail · fp4 | $4.50 | $22.50 | 100% | Read |
| Fireworks | $4.50 | $22.50 | 88% | Read |
| Fireworks | $4.50 | $22.50 | 94% | Read |
Based on OpenRouter and Vercel AI Gateway data. Confirm terms with the provider before sending sensitive data.
What it can do
- Reads: text, image, video, pdf
- Writes: text
- Weights: Open weights, restricted
- Reasoning: Yes
- Tool calling: Yes
- Longest reply: 128K tokens
- Repeated text is billed at $0.182 per 1M (prompt caching)
Compare with
Appears in
Estimate your monthly cost with Kimi K3
Sources and licenses
- OpenRouter: Prices, context windows, modalities and the model list come from the OpenRouter public API. Data: OpenRouter.
- Vercel AI Gateway: The no-training flag comes from the Vercel AI Gateway model list.
- models.dev: Lab launch dates, open-weights flags and knowledge cutoffs come from models.dev. Copyright (c) 2025 models.dev, MIT License.
- Epoch AI: Independent benchmark scores and the Epoch Capabilities Index (ECI) come from Epoch AI's Benchmarking Hub, used under CC BY 4.0. Epoch AI, Capabilities & benchmarking, epoch.ai/benchmarks. Scores Epoch marks as external keep their original license, and the source is named on each score.
- LMArena: Arena Elo (text, webdev and vision) comes from the LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset), CC BY 4.0, leaderboard date 2026-10-28.
- Lab-reported scores, when an editor has added them, are the lab's own numbers. Each one links to the source and is labeled lab-reported. They are not used to rank the lists.
- What a score measures: the benchmark guide.
