Kimi K2.5

Tool callingStructured outputReasoningImage inputVideo input

Hugging Face Launch post Website Docs

Earlier Kimi frontier model for long-context agents, coding, and multimodal work

Context: 262K
Released: 9mo ago
Parameters: —
Input / 1M tokens: $0.50
Output / 1M tokens: $2.50
Cached input / 1M tokens: $0.15

Sources:models.devOpenRouter

Benchmarks

DeepSWE is the primary coding benchmark where an exact model evaluation exists. Other sourced evaluations appear below; missing scores are not zero.

Not evaluated in DeepSWE 1.1 as of 2026-09-22. Results are only shown for exact model versions in the official dataset. Official leaderboard.

Other published benchmarks

Evaluation	Reported score and source
ARC-AGI-2	11.8% Third-party evaluator · kimi-k2.5 External evaluation via Epoch AI Original-source terms apply
OSWorld	63.3% OS World Website · kimi-k2.5 Agent: Kimi-K2.5 OS World Website Original-source terms apply
SimpleQA Verified	34.3% Epoch AI · kimi-k2.5 Epoch AI evaluation
SWE-Bench verified	73.8% Epoch AI · kimi-k2.5 Epoch AI evaluation
Terminal Bench	43.2% https://www.tbench.ai/leaderboard/terminal-bench/2.0 · kimi-k2.5 Agent: Terminus 2 https://www.tbench.ai/leaderboard/terminal-bench/2.0 Original-source terms apply

Benchmark compilation: Epoch AI, Capabilities & benchmarking (CC BY attribution). Scores are converted to percentages and model names matched to the Vertical Tabs catalog. Individual evaluations are credited to their original sources. External results may carry separate licences and evaluation configurations; do not interpret scores from different harnesses as directly comparable.

Specifications

Developer: Moonshot AI
Model family: Kimi
Type: Language model
Predecessor: —
Successor: —
Released: January 1, 2026
Model ID: moonshotai/kimi-k2.5
Openness: Open weights
License: —
Parameters: —
Input modalities: text, image
Output modalities: text
Modality: text+image->text
Tokenizer: Other
Max output tokens: 236K
Knowledge cutoff: 2025-01
Hugging Face: moonshotai/Kimi-K2.5
Source freshness: Metadata synced Oct 8, 2026, 10:17 PM UTC

Additional source evidence

Source model: moonshotai/kimi-k2.5
Source family: kimi-k2
Last updated: 2026-01
Knowledge cutoff: 2025-01
Model context: 262K
Model output limit: 262K
Open weights: Yes
File attachments: Yes
Reasoning: Yes
Tool calling: Yes
Structured output: Yes
Temperature control: No

Downloadable weights

Hugging Face

Published benchmarks

Benchmark	Score	Evaluation details	Evidence
SWE-Bench Verified	70.8 · resolved	—	Source
SWE-Atlas Codebase QnA	13.1 · score	Mini-SWE-Agent	Source
SWE-Atlas Refactoring	20.95 · score	Mini-SWE-Agent	Source
SWE-Atlas Test Writing	25.77 · score	Mini-SWE-Agent	Source

Complete source record (JSON)

{
  "id": "moonshotai/kimi-k2.5",
  "name": "Kimi K2.5",
  "description": "Earlier Kimi frontier model for long-context agents, coding, and multimodal work",
  "family": "kimi-k2",
  "last_updated": "2026-01",
  "knowledge": "2025-01",
  "attachment": true,
  "reasoning": true,
  "tool_call": true,
  "structured_output": true,
  "temperature": false,
  "modalities": {
    "input": [
      "text",
      "image",
      "video"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 262144,
    "output": 262144
  },
  "benchmarks": [
    {
      "name": "SWE-Bench Verified",
      "score": 70.8,
      "source": "https://www.swebench.com/",
      "metric": "resolved"
    },
    {
      "name": "SWE-Atlas Codebase QnA",
      "score": 13.1,
      "source": "https://labs.scale.com/leaderboard/sweatlas-qna",
      "metric": "score",
      "harness": "Mini-SWE-Agent"
    },
    {
      "name": "SWE-Atlas Refactoring",
      "score": 20.95,
      "source": "https://labs.scale.com/leaderboard/sweatlas-refactoring",
      "metric": "score",
      "harness": "Mini-SWE-Agent"
    },
    {
      "name": "SWE-Atlas Test Writing",
      "score": 25.77,
      "source": "https://labs.scale.com/leaderboard/sweatlas-tw",
      "metric": "score",
      "harness": "Mini-SWE-Agent"
    }
  ],
  "release_date": "2026-01",
  "open_weights": true,
  "weights": [
    {
      "label": "Hugging Face",
      "url": "https://huggingface.co/moonshotai/Kimi-K2.5"
    }
  ]
}

Infrastructure

Provider availability

All providers

Provider	Endpoint	Context	Input /M tokens	Output /M tokens	Cached input /M tokens	Cache write /M tokens	Uptime 1d	Data handling
AtlasCloud	atlas-cloud/int4	262K	$0.49	$2.50	$0.20	—	97.08%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
SiliconFlow	siliconflow/int4	262K	$0.50	$2.50	$0.15	—	99.93%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Venice	venice	256K	$0.53	$3.32	$0.21	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Novita	novita	262K	$0.57	$2.85	$0.10	—	99.06%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Amazon Bedrock	amazon-bedrock/us-east-2	262K	$0.60	$3.00	—	—	99.70%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15

Provider data:OpenRouter

Deployment graph

Can I run this?

Vertical Tabs only shows compatibility when it has structured evidence tying this exact model to hardware, runtime, quantization, memory use, and observed performance.

Browse hardware