Scheduled for retirement on Jan 26, 2027. Still available at the prices below until then.
Context
262K
Max output
262K
Serving providers
11
Cheapest input
$0.45 /1M
On-demand listed rates, OpenRouter routed quotes and Vercel reseller quotes. Enterprise commitments (committed-use discounts, provisioned throughput, negotiated rates) can be materially lower.
Benchmarks
independent evaluations · source: ARC Prize, Artificial Analysis and Hugging Face leaderboardsSpeed & latency
Output speed
60 t/s
tokens / second
Time to first token
1.60s
latency
Headline indices
Individual benchmarks · click a row for effort variants
| Benchmark | Best score | Effort | Cost / task | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AIME 2026 | 96% | default | — | |||||||||||||||||
| APEX Agents | 12% | default | — | |||||||||||||||||
| ARC-AGI-2 | 12% | default | $0.28 | |||||||||||||||||
| CritPt | 3% | default | — | |||||||||||||||||
| ||||||||||||||||||||
| EvalComputeProxyFrozen Aug 18, 2026 | 648.1 | default | — | |||||||||||||||||
| ||||||||||||||||||||
| GDPval v2Frozen Aug 18, 2026 | 1006.1 | default | — | |||||||||||||||||
| ||||||||||||||||||||
| GPQA Diamond | 88% | default | — | |||||||||||||||||
| ||||||||||||||||||||
| Humanity's Last Exam | 31% | default | — | |||||||||||||||||
| ||||||||||||||||||||
| IFBench | 70% | default | — | |||||||||||||||||
| ||||||||||||||||||||
| Long-Context Reasoning | 78% | default | — | |||||||||||||||||
| ||||||||||||||||||||
| MMMU-Pro | 75% | default | — | |||||||||||||||||
| ||||||||||||||||||||
| Omniscience | -7.3 | default | — | |||||||||||||||||
| ||||||||||||||||||||
| Omniscience Accuracy | 0.4 | default | — | |||||||||||||||||
| ||||||||||||||||||||
| Omniscience Non Hallucination | 0.5 | reasoning: false | — | |||||||||||||||||
| ||||||||||||||||||||
| SciCodeArtificial Analysis did not report this score in its latest run on Oct 14, 2026 | 49% | — | — | |||||||||||||||||
| ||||||||||||||||||||
| SWE-bench Pro | 51% | default | — | |||||||||||||||||
| SWE-bench Verified | 71% | default | — | |||||||||||||||||
| TerminalBench Hard | 35% | default | — | |||||||||||||||||
| ||||||||||||||||||||
| TerminalBench v2.1 | 46% | default | — | |||||||||||||||||
| ||||||||||||||||||||
| τ-bench Banking | 14% | default | — | |||||||||||||||||
| τ²-bench | 96% | default | — | |||||||||||||||||
| ||||||||||||||||||||
Artificial Analysis last published scores for this benchmark on Sep 9, 2026. Scores here are frozen at that date, and models released since are not on it.
Artificial Analysis last published scores for this benchmark on Aug 18, 2026. Scores here are frozen at that date, and models released since are not on it.
Pricing by serving provider
standard tier · per 1M tokens · click a provider to expandInput pricing runs from $0.45 to $0.60 per 1M tokens across 11 providers, so the dearest route costs 33% more than the cheapest for the same model.
| Serving provider | Input /1M | Output /1M | Endpoints | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| DeepInfraCheapest | $0.45 | $2.25 | 1 | |||||||||||||
Standarddeepinfra/moonshotai/Kimi-K2.5
| ||||||||||||||||
| OpenRouter | $0.50 | $2.50 | 1 | |||||||||||||
Standardmoonshotai/kimi-k2.5
| ||||||||||||||||
| Together AI | $0.50 | $2.80 | 1 | |||||||||||||
Standardtogether_ai/moonshotai/Kimi-K2.5
| ||||||||||||||||
| Azure | $0.60 | $3.00 | 1 | |||||||||||||
Standardazure_ai/kimi-k2.5
| ||||||||||||||||
| Baseten | $0.60 | $3.00 | 1 | |||||||||||||
Standardbaseten/moonshotai/Kimi-K2.5
| ||||||||||||||||
| Amazon Bedrock | $0.60 | $3.00 | 1 | |||||||||||||
Standardbedrock/moonshotai.kimi-k2.5
| ||||||||||||||||
| Fireworks AI | $0.60 | $3.00 | 1 | |||||||||||||
Standardfireworks_ai/accounts/fireworks/models/kimi-k2p5
| ||||||||||||||||
| Moonshot | $0.60 | $3.00 | 1 | |||||||||||||
Standardmoonshot/kimi-k2.5
| ||||||||||||||||
| Azure is dearest at $0.60 / $3.00 | ||||||||||||||||
Price history
listed seller prices, input and output per 1M tokens Input Output
Input down 25% since first tracked
Cost calculator
estimate your monthly spend on this model$270
estimated / month
Model IDs
copy the exact identifier for your platformdeepinfra/moonshotai/Kimi-K2.5together_ai/moonshotai/Kimi-K2.5moonshotai/kimi-k2.5fireworks_ai/accounts/fireworks/models/kimi-k2p5bedrock/moonshotai.kimi-k2.5baseten/moonshotai/Kimi-K2.5moonshot/kimi-k2.5wandb/moonshotai/Kimi-K2.5novita/moonshotai/kimi-k2.5azure_ai/kimi-k2.5moonshotai/kimi-k2.5
Frequently asked questions
Kimi K2.5 pricing, context and availabilityIs Kimi K2.5 still available?
Yes. Kimi K2.5 is scheduled for retirement on 2027-01-26.
How much does Kimi K2.5 cost?
Kimi K2.5 costs $0.45 per 1M input tokens from DeepInfra, and $2.25 per 1M output tokens from DeepInfra. Across 11 serving providers, input prices range from $0.45 to $0.60 per 1M tokens.
What is the context window of Kimi K2.5?
Kimi K2.5 accepts up to 262K tokens of context and can return up to 262K output tokens.
Which providers serve Kimi K2.5?
Kimi K2.5 is available from 11 serving providers, each with its own pricing and model ID. DeepInfra is currently the cheapest.
What can Kimi K2.5 do?
Kimi K2.5 supports image input (vision), tool use, reasoning, prompt caching and structured output.
Other Moonshot models
compare pricing across the Moonshot lineup Listed prices from LiteLLM, OpenRouter and Vercel · last checked Oct 14, 2026
Every weekday
AI moves fast. Here's your debrief.
News, analysis, tools, and more.
For people who build with AI