LLM Reference

Kimi K2.6

Released
2026-04-20
Last refreshed
2026-06-30
Status
Researched 18d ago
Open sourceCommercial use: permittedMultimodalCodingRAGAgentsLong contextVisionClassificationJSON / Tool use

Kimi K2.6 is worth evaluating for coding, rag, and agents when its provider route and context window match the workload.

Use it for

  • Teams evaluating coding, rag, and agents
  • Workloads that can use a 262k context window
  • Buyers comparing 4 tracked provider routes

Do not use it for

  • Workloads where another current model has stronger sourced task evidence
Specifications
Family
Kimi K2
Released
2026-04-20
Context
262k
Max output
65,536
Parameters
1T
Architecture
Mixture of Experts
Knowledge cutoff
2025-04
Specialization
code
Openness
Open source
License
MITOSI-approvedCommercial use: permitted
Created by

Lossless long-context AI innovation

Beijing, China
Founded 2023
Website
Pricing
Output / 1M
$3.40
Input / 1M
$0.800

Cheapest of 9 routes · Novita AI

About

Kimi K2.6 is Moonshot AI's multimodal agentic coding model, released April 20 2026 under a Modified MIT license. Built on a 1-trillion-parameter MoE architecture (32B active, 384 experts with 8 selected per token plus 1 shared expert, 61 layers), it features a 262K context window and up to 65,536 output tokens. Supports native image and video inputs (screenshots, PDFs, spreadsheets). Designed for long-horizon coding with agent swarms of up to 300 sub-agents and 4,000 coordinated steps; Moonshot AI cites 200–300 sequential tool calls without task drift. Key benchmarks: SWE-bench Verified 80.2%, SWE-bench Pro 58.6%, LiveCodeBench v6 89.6%, GPQA Diamond 90.5%, Terminal-Bench 2.0 66.7%. Chatbot Arena Elo 1454 (2026-04-28 snapshot).

Kimi K2.6 is an open-source model in the Kimi K2 family. The structured metadata tracks a 262k-token context window, multimodal input, reasoning, function calling, tool use, and structured outputs. This page tracks provider routes through Cloudflare Workers AI, NVIDIA NIM, Moonshot AI Kimi, and 6 more, with the cheapest tracked route listed at $0.73 input and $3.49 output per 1M tokens. Headline tracked benchmarks include SWE-bench Verified 80.2, Chatbot Arena 1462.0, and Google-Proof Q&A 90.5.

Top use-case fit: coding, agents, and build tasks

Coding

Q/$ D

4 relevant benchmarks in the decision map.

RAG

Included by capability and metadata signals in the decision map.

Agents

Q/$ C

1 relevant benchmark in the decision map.

Provider price ladder

Compare all 9

Compare API pricing across 4 providers for input and output tokens, batch, and cached reads when available.

ProviderInput / 1MOutput / 1MRoute
Novita AI$0.800$3.40
Serverless
OpenRouter$0.730$3.49
Serverless
Cloudflare Workers AI$0.950$4.00
Serverless
Fireworks AI$0.950$4.00
Serverless

Available via routers & gateways(7)

Capabilities

VisionMultimodalReasoningFunction CallingTool UseStructured OutputsPrompt Caching

Benchmark peer barsfor Coding

Benchmark scores(17)

Scores are benchmark-specific and are direction-aware: the same numeric gap can mean very different outcomes across suites. Use the leaderboard context and this model's provider route to decide whether the winning margin is meaningful for your workload.
BenchmarkScoreVersionEvaluationSource
SWE-bench Verified80.2SWE-bench VerifiedObserved 2026-04-24Source
Chatbot Arena1462.0Observed 2026-05-17Source
Google-Proof Q&A90.5GPQA Diamond (accuracy)Observed 2026-06-07Source
MMLU PRO84.6From Kimi K2 (accuracy)Observed 2026-06-07Source
HumanEval92.0Observed 2026-04-29Source
SWE-bench Pro58.6SWE-bench Pro (resolved%)Observed 2026-06-07Source
Instruction-Following Evaluation89.8Observed 2026-04-29Source
MMMU Pro80.1LLM-Stats aggregatorObserved 2026-06-07Source
LiveCodeBench89.6v6Observed 2026-04-20Source
Terminal-Bench 2.066.7Terminal-Bench 2.0 (accuracy%)Observed 2026-06-07Source
SWE-bench Multilingual76.7Observed 2026-04-20Source
AIME 202696.4AIME 2026 (accuracy)Observed 2026-06-07Source
BrowseComp83.2BrowseComp (accuracy%)Observed 2026-06-07Source
Humanity's Last Exam34.7HLE-Full without tools (accuracy)Observed 2026-06-07Source
MCP-Atlas55.9MCP-Atlas (accuracy%)Observed 2026-06-07Source
CursorBench47.6CursorBench 3.1Observed 2026-06-30
Configuration: Kimi 2.6 (single reported configuration)
Harness: CursorBench 3.1
Evaluator: Cursor
Confidence: confirmed
Notes: Cursor published one CursorBench 3.1 configuration for this model; no cross-effort selection was needed.
Source
GeneBench-Pro4.4xhighObserved 2026-06-30Source

Compare Kimi K2.6 to GPT-5.5 head-to-head for benchmarks, pricing, and full specs side by side. On NVIDIA NIM, see pricing and specs or the setup guide (catalog ID moonshotai/kimi-k2.6).

Migration checks

No linked migration route is available for this model yet.

Compare Kimi K2.6 with other models

Show all 32 popular comparisonssorted by 7-day search impressions

Frequently asked questions

What is the context window of Kimi K2.6?

Kimi K2.6 has a context window of 262k tokens.

What is the max output of Kimi K2.6?

Kimi K2.6 can generate up to 65,536 output tokens.

How much does Kimi K2.6 cost?

Kimi K2.6 pricing ranges from $0.73/1M to $1.20/1M input tokens depending on the provider.

When was Kimi K2.6 released?

Kimi K2.6 was released on 2026-04-20.

Which providers offer Kimi K2.6?

Kimi K2.6 is available from 9 providers: Cloudflare Workers AI, NVIDIA NIM, Moonshot AI Kimi, Fireworks AI, OpenRouter, Microsoft Foundry, Vercel AI Gateway, Novita AI, Together AI.

What benchmarks has Kimi K2.6 been tested on?

Kimi K2.6 has been evaluated on 17 benchmarks, including SWE-bench Verified, Chatbot Arena, Google-Proof Q&A, MMLU PRO, HumanEval.