Moonshot AI·moonshotai/kimi-k2.5
Kimi K2.5 is a language model published by Moonshot AI. 10 OpenRouter endpoints are published in the latest snapshot. It accepts text, image input and returns text. The published context limit is 262K tokens. The catalogue declares support for reasoning, tool use, structured outputs.
Superseded by Kimi K3. It no longer holds a current-board position; its all-versions archive rank is #36.
Providers
Uptime measured over the last 30 minutesHugging Face Inference Providers
Exact repository identityProvider economics and telemetry published by the Hugging Face Router. These are operational signals, not benchmark results.
Reasoning efforts
Best supported result can inform ModelCapPublished Max, X-High, High and other configurations for this model. These are sourced scores, not head-to-head essays.
Related community configurations
Human preference ratingReported evaluation leads (81)
Excluded from scoreStructured rows reported in this exact model repository at the pinned revision below. They are useful research leads, but remain quarantined until dataset revision, harness, configuration and identity pass ModelCap’s independent admission review.
Showing 20 of 81 rows.
Other versions (4)
ModelCap Index · All-versions archive
Methodology- Public rank basis
- Measured (measured)
- Index support
- 91.4% · strong
- Index interval
- 78.1–83.5
- Rank posterior
- Not available
- Top 5 / Top 10 probability
- Not available / Not available
- Posterior as of
- Snapshot timestamp unavailable
- Identity binding
- Exact catalogue product · moonshotai/moonshotai/kimi-k2.5 · aggregates disclosed benchmark configurations · endpoint configuration not claimed · artifact metadata moonshotai/Kimi-K2.5@4d01dfe0332d63057c186e0b262165819efb6611 (not the evaluation revision)
- Direct benchmark observations
- 4
- Breadth-adjusted BES point estimate
- 80.8
BES normalizes admitted public benchmark evidence for measured Index rows. Its legacy score and observables prior do not define the public language rank.
- Observed capability
- 80.8
- Evidence status
- confirmed · 95% mass
- General preference
- 82.9
- Coding
- 78.9
- Agents & tools
- —
- Reasoning
- 45.9
- Evidence breadth
- 95%
- Evidence coverage
- 91%
- Usage (OpenRouter popularity)
- 30.6
- Liquidity (providers × uptime)
- 98.6
- Open reach (HF downloads)
- —
- Surface (context / tools / modalities)
- 90.3
- Freshness
- 48.0
- · OpenRouter popularity #52
- · 10 live providers
- · Catalogue: context 262144, reasoning, tools, modalities Text/Image
- · First seen 2026-01-27
These adoption and deployment observations remain context only. They do not change this model's ModelCap Index score or rank.
- Arena Overall · Thinking82.9
- SWE-bench bash-only · High74.6
- Arena Coding · Thinking84.1
- ARC-AGI-2 Verified45.9
The bounded 0–100 capability score blends each source’s competitive placement with its published achievement, then combines capability families. Evidence breadth adds a modest uncertainty adjustment; price and popularity are not part of benchmark-led rank.
Deployment readiness
Metadata index- Artifact reproducibility
- 85.0
- Access & legal clarity
- 67.5
- Deployability
- 97.3
- Evaluation provenance
- 82.4
Missing: artifactReproducibility.base-lineage, accessLegalClarity.resolvable-license-terms, evaluationProvenance.evaluation-dataset-revisions.
A metadata completeness and deployability index—not a safety certification, quality grade, or production-readiness claim.
Market activity
Venue coverage- Market Gravity
- 51.4
- OpenRouter weekly popularity
- #52 of 298
Market Gravity uses OpenRouter’s full-catalogue popularity order. Sparse Vercel top-ten observations appear only when published and do not affect the score.
Arena preference
Arena has published current community results for Thinking. No default-configuration result is published, so the raw Arena field stays configuration-specific. The strongest supported result can still inform ModelCap Score with reduced confidence.
Current board through 6 Aug 2026, 13:00 UTC
Market Gravity breakdown
- Usage55%
- 30.6
- Liquidity25%
- 98.6
- Open reach15%
- 50.0
- Freshness5%
- 48.0
OpenRouter weekly popularity across the full model catalogue.
Independent providers versus the model's open or closed cohort, weighted by uptime.
Hugging Face 30-day downloads for open models; neutral for closed models.
Time since first public availability, on a six-month half-life.
Specification
- Context window
- 262K tokens
- Max output
- 262K tokens
- Inputs
- Text, Image
- Outputs
- Text
- Tokenizer
- Other
- Cached input tokens
- $0.095 / 1M
- First seen on OpenRouter
- 27 Jan 2026
Capabilities
- Supported: Reasoning
- Supported: Tool use
- Supported: Structured output
- Supported: Response format
- Not supported: Moderated
Weights & access
Restricted license. Downloadable weights use a model-specific or use-restricted license. Review the linked license before use, redistribution, or commercial deployment.
- Access
- Restricted license
- Licence
- modified-mit
- Pinned revision
- 4d01dfe0332d
- Parameters
- 1058.6B
- Architecture
- KimiK25ForConditionalGeneration
- Model type
- kimi_k25
- Hugging Face downloads (30d)
- 919K
- Downloads (all time)
- 15.1M
- Likes
- 3K
- Repository
- moonshotai/Kimi-K2.5
- Repository updated
- 30 Apr 2026, 03:56 UTC