Kimi K2.5
moonshot/kimi-k2.5- Moonshot AI
- Family · Kimi K2
- Released 27 Jan 2026
- Open weights
- Text · Image → Text
- Tools · JSON · Vision
Spec & list price
Context250K
Max output236K
Input / 1M$0.50
Output / 1M$2.505× input
Scores · standing
Intelligence23.5#108/489
Coding46.8#74/196
Usage
Tokens / day2.6B▼ 23% 7d
Usage rank#145
Requests / day129.6K
Tool calls / req0.14
Per request · 30 days
Avg input11.8K0.8× median
Avg output9491.4× median
Cache hit64%about median
Reasoning share68%1.2× median
Tool-call failures2.3%1.2× median
Consumption
Peak 7.5B 5 OctLow 1.8B 1 NovRank today #145
News & trends
2 mentions this weekthis model · videos, posts, articles · 30 days
Trending · 14 days
- Kimi k2.57 upvotes
- Stuck between GLM 4.7, 4.6, and Kimi K2.5. Also, which provider should I use between OpenRouter and DeepInfra?49 upvotes
- #36 Clawdbot (OpenClaw), Moltbook, Variant AI, Pencil.dev, Kimi K2.5, Project Genie8 views
News · latest from Moonshot AI
- ReplaySSM: Rethinking State Storage for Faster Hybrid-Model DecodingNVIDIA · 12 Oct
- Together Link: open models in the harness you already use. Start with one command today.Together AI · 5 Oct
- Kimi K3: the complete developer guideTogether AI · 1 Aug
- Together AI announces strategic partnership with Moonshot AI to natively serve Kimi modelsTogether AI · 29 Jul
Benchmarks
bar = share of the top score · Benchmark data: Artificial AnalysisSpeed · median across 5 providers, latest day
Output speed#254/33444 tok/s1.5× slower than median
Fastest provider87 tok/sbest of 5 providers
First token#1621.0 sabout median
#18 cheapest of 33 with a similar score
Value
Intelligence Index
Blended price per 1M tokens (log)
Family progress
Intelligence Index
- Family releases27.0
Technical details
- Tokenizer
- Other
- License
- other
- Parameters
- 1T
- Moderation
- none
- Downloads · likes
- 407.9K · 2.9K
Request parameters 19
- tools
- tool_choice
- response_format
- structured_outputs
- reasoning
- include_reasoning
- temperature
- top_p
- top_k
- min_p
+9 more
- max_tokens
- stop
- seed
- frequency_penalty
- logit_bias
- logprobs
- presence_penalty
- repetition_penalty
- top_logprobs
Interesting facts
- 2.2× the output room of similar models (235.9K against a 106.5K median).
- 2nd of 5 Kimi K2 models by tokens last month.
When to use
Best for
- Tool-using agents and multi-turn workflows
- Science and graduate-level question answering
- Long document reasoning
- Image understanding in chat apps
- Self-hosted deployments on open weights
Think twice for
- Terminal-heavy coding agents
- Video input
- New projects (usage is declining)
Research papers
- Evolving to Match Capabilities: Improving Harness Self-Evolution via Skill Adaptation10 Oct
- From Pixel to Coding: Evaluating the Figure Reproduction Capabilities of MLLMs7 Oct
- One Step at a Time: Trading LLM Autonomy for Process Predictability6 Oct
- CTE-Bench: Counterfactual Trace Evaluation for Stateful Software Simulators29 Sep
- The Complexity Kink: A Prompt-Side Structural Complexity Index for Code-Generation Reliability17 Sep
- Big Enough to Break Out: Tracking the Rising Capability of LLM Penetration-Testing Agents9 Sep
- Fortunate Recall: Ontology-Driven Memory Lifecycle Management for Persistent Coherence in LLMs9 Sep
- VDiff-Bench: A Challenging Benchmark for Fine-Grained Image Difference Identification5 Sep
Social
Most watched and discussed · 90 days
- #36 Clawdbot (OpenClaw), Moltbook, Variant AI, Pencil.dev, Kimi K2.5, Project GenieCmd+AI · 8 views · 20 Oct
Same family
Kimi K2- 22.0
- 27.0
- 12.7
- 15.3
Alternatives
other vendors · similar capability- 26.3
- 21.8
- T25.0
- 21.5
Using Kimi K2.5 at volume?
Sign in to ask for a volume discount. You will continue in the OSMEXO dashboard.