Kimi-K2.6

moonshotai Kimi (Modified MIT) (modified-mit) out 6 months ago

1.03T parameters (8 of 384 experts per token), images and text in, text out, 256K tokens of context. 18 providers serve it from $0.45 per million input tokens.

Verified 2026-10-25 Rank #26 in LLMs · score 115.5 method v1 Sources 5 Downloads 30d 348.2k · #199 of 619 tracked

Changed this week: 1 new quantized build on the Hub (45 in all).2026-10-24

Can I run it?616GB at INT4Estimateneeds a multi-GPU nodemethodCommunity: Cerebras Brings Trillion Parameter Inference to Enterprises with Kimi K2.6 →

What does it cost?$0.45/1M incheapest of 17Baidufp42026-10-24

Can I use it?Kimi (Modified MIT) (modified-mit)MIT plus a display-the-model-name attribution requirement for products above 100M MAU or US$20M monthly revenue; the exact wording is in each repo's LICENSElicence file2026-10-24

Do people use it?348.2kdownloads, 30 days#199 of 619 we track · top 32%+5 likes, downloads −9% this week1,636 likesHugging Face2026-10-24

Worth reading

Picked by rule from 24 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Kimi K2.6 optimizes model inference in Zig, a highly niche programming language, to demonstrate exceptional out-of-distribution generalization.

From Kimi K2.6: Advancing open-source coding · kimi.com · 2026-04-20

The lab's own post

Kimi K2.6: Advancing open-source coding

Kimi K2.6 is an open-source model released for advanced coding, long-horizon execution, and agent swarm capabilities.

Kimi K2.6 optimizes model inference in Zig, a highly niche programming language, to demonstrate exceptional out-of-distribution generalization.
The model reconfigured the exchange-core thread topology from 4ME+2RE to 2ME+1RE to extract a 185% medium throughput leap.

kimi.com 2026-04-20 710 HN pts · 372 comments 2.3k words

Most discussed · 380 HN pts

Kimi K2.6 just beat Claude, GPT-5.5, and Gemini in a coding challenge

Kimi K2.6 won the Word Gem Puzzle challenge with 22 match points and a 7-1-0 record, beating GPT-5.5 and Claude Opus 4.7.

Nvidia's Nemotron Super 3 produced code with a syntax error that prevented connection to the game server.
Muse Spark claimed every visible word regardless of length, resulting in a cumulative score of −15,309.

thinkpol.ca 2026-05-03 380 HN pts · 219 comments 1.4k words

Latest · 3 months ago

US lawmakers investigate DoorDash's use of Moonshot AI's Kimi K2.6 model

scmp.com 2026-08-01 11 HN pts · 4 comments

24 best of 24 tracked

The 5 picks above, plus these 19. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

ArticleKimi K2.6-code-preview is now availablekimi.com · 2026-04-13 · 14 HN pts · official
DiscussionKimi K2.6: Advancing Open-Source Codingtwitter.com · 2026-04-20 · 40 HN pts
ArticleBerget AI announces Berget Code for European teams powered by Kimi K2.6berget.ai · 2026-05-13 · 5 HN pts
ArticleOverview of Kimi K2.6 Modelplatform.kimi.ai · 2026-04-21 · 2 HN pts · official
CodeClaude-Code-Proxy – Use Kimi K2.6 and OpenAI Subscriptions in Claude Codegithub.com · 2026-05-04 · 2 HN pts
DiscussionShow HN: Kimiflare, Kimi K2.6 as a Coding Agent on Cloudflare Workers AInews.ycombinator.com · 2026-05-12
ArticleDeepSeek V4 Pro and Flash vs. Claude Opus 4.7 and Kimi K2.6blog.kilo.ai · 2026-05-15
ArticleCerebras Brings Trillion Parameter Inference to Enterprises with Kimi K2.6cerebras.ai · 2026-05-20
ArticleWe Gave Claude Opus 4.7 and Kimi K2.6 the Same Workflow Orchestration Specblog.kilo.ai · 2026-04-22
ArticleKimi K2.6 Deep Dive: What Makes This Model Different?aiindigo.com · 2026-06-19
ArticleKimi K2.6 Deep Dive: Is it the New King of Open Weights?aiindigo.com · 2026-06-14
ArticleKimi K2.6, reviewed — benchrbenchr.org · 2026-05-30
ArticleKimi K2.6 Review: Moonshot AI's Open-Weight Trillion-Parameter Agent — Agent Swarm, MLA Attention, 80.2% SWE-Bench — ChatForestchatforest.com · 2026-05-13
ArticleKimi K2.6: Open-Source Just Beat GPT-5.5 at Codingbuildfastwithai.com · 2026-05-04
ArticleKimi K2.6 Review (8.5/10): 1T Open MoE Beats Opus on Cost | ThePlanetTools.aitheplanettools.ai · 2026-04-29
ArticleKimi-K2.6: release, architecture & GPU deployment | AIInfraInterviewsaiinfrainterviews.com · undated
ArticleKimi K2.6 — AI Pro Playbook | AI Pro Playbookaiproplaybook.com · undated
ArticleKimi K2.6 with Strix: a quick testtheartificialq.github.io · 2026-04-21
ArticleKimi K2.6 Developer Guide: Benchmarks, API & Agent Swarm | Lushbinarylushbinary.com · 2026-04-20

Show all 19

Where to run it

Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.

Prices for the same weights vary 2×

17 providers with a live price, from $0.45 to $1.09 per million input tokens; the median is $0.75. One dot per provider, cheapest in teal.

#	Cheapest 5	In /1M	Out /1M	Context	Quant	Seen
1	Baidu	$0.45	$1.89	256K	fp4	Oct 24
2	Chutes	$0.45	$1.95	256K	int4	Oct 24
3	DigitalOcean	$0.57	$2.40	256K	unknown	Oct 24
4	StreamLake	$0.60	$2.52	250K	fp8	Oct 24
5	Inceptron	$0.60	$3.20	256K	int4	Oct 24

All 18 providers →

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

Widely served (18 API providers); 45 quantized builds for local use.

As a service18AtlasCloud, Baidu, Chutes, Cloudflare +14
On own hardware45quantized builds
Built on2918 fine-tunes · 9 adapters · 2 merges
In apps100+Hugging Face Spaces

Top Spaces running it: akhaliq/Kimi-K2.6 ♥ 15 · Quanyu001/trace2env_demo ♥ 3 · shrinusn77/kimi-k2.6-chatbot ♥ 2 · xyanglu/resume-agent-chatbot ♥ 1

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Needs an 8×80 GB node even at INT4

About 616 GB at INT4, 1.23 TB at FP8, 2.46 TB at BF16: weights only, so leave room for the KV cache. Every expert must be resident, so total parameters count.

Estimate params × bytes per param × 1.2 · how we estimate

Day-one receipts

Claim-matched tests, graded by code. Receipts, not a score.

Speed run● pass

{
  "providers": [
    {
      "provider": "parasail",
      "endpoint_model": "parasail-kimi-k26",
      "source_url": "https://api.parasail.io/v1/chat/completions",
      "ttft_ms_median": 6905,
      "ttft_any_ms_median": 561,
      "tokens_per_s_median": 132,
      "runs": [
        {
          "ttft_s": 6.905,
          "ttft_any_s": 0.66,
          "tokens_per_s": 177.6,
          "latency_s": 7.61,
          "tokens_out": 1352,
          "reasoning_chars": 5792,
          "finish": "stop"
        },
        {
          "ttft_s": 12.411,
          "ttft_any_s": 0.561,
          "tokens_per_s": 55.7,
          "latency_s": 13.31,
          "tokens_out": 741,
          "reasoning_chars": 2716,
          "finish": "stop"
        },
        {
          "ttft_s": 5.103,
          "ttft_any_s": 0.469,
          "tokens_per_s": 132,
          "latency_s": 5.91,
          "tokens_out": 780,
          "reasoning_chars": 2990,
          "finish": "stop"
        }
      ]
    }
  ],
  "note": "Single client, wall clock; tok/s counts every output token including hidden reasoning."
}

Three-tool errand● pass

{
  "calls": [
    {
      "name": "get_weather",
      "args": {
        "city": "Lisbon",
        "date": "2026-10-14"
      }
    },
    {
      "name": "check_calendar",
      "args": {
        "date": "2026-10-14"
      }
    },
    {
      "name": "book_table",
      "args": {
        "restaurant": "Cervejaria Ramiro",
        "date": "2026-10-14",
        "time": "19:30",
        "party_size": 2
      }
    }
  ],
  "checks": {
    "get_weather": true,
    "check_calendar": true,
    "book_table": true
  },
  "final": "The weather in Lisbon on Wednesday, 2026-10-14 looks great — clear skies with only a 5% chance of rain.\n\nYour free evening slot that day is **19:30–21:30**, so I’ve booked a table for 2 at **Cervejaria Ramiro** for **19:30**.\n\n**Confirmation code: TBL-4471**",
  "fixture": "receipts.errand.v1"
}

The story so far

Weights, every dated article, and repo commits on one rail.

Most coverage came in the first two weeks

Teal is the weights (2026-04-14), amber triangles are 9 dated articles (each links to it), grey ticks are repo commits.

What people built on it

74 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

74 derivatives on the Hub, most of them quantized

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

GGUF unsloth/Kimi-K2.6-GGUF for llama.cpp, Ollama, LM Studio 985.1k downloads MLX mlx-community/Kimi-K2.6-mlx-DQ3_K_M-q8 for Apple silicon 369k downloads NVFP4 nvidia/Kimi-K2.6-NVFP4 for vLLM, SGLang 2.3M downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

23 rows

Benchmarks 23 rows

measured third-party leaderboard rowauthor-reported number from the model card or eval files (23)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.

Benchmark	Value	Setting / notes	Reported by	Date
author-reported APEX-Agents	27.9	Agentic	model author	2026-10-07
author-reported BrowseComp	83.2	Agentic	model author	2026-10-07
author-reported Claw Eval (pass@3)	80.9	Agentic	model author	2026-10-07
author-reported Claw Eval (pass^3)	62.3	Agentic	model author	2026-10-07
author-reported HLE-Full	34.7	Reasoning & Knowledge	model author	2026-10-07
author-reported LiveCodeBench · Write code with agents	89.6	Coding	model author	2026-10-07
author-reported MCPMark	55.9	Agentic	model author	2026-10-07
author-reported OJBench (python)	60.6	Coding	model author	2026-10-07
author-reported OSWorld-Verified	73.1	Agentic	model author	2026-10-07
author-reported SWE-bench Multilingual · Write code with agents	76.7	Coding	model author	2026-10-07
author-reported SWE-bench Pro · Write code with agents	58.6	Coding	model author	2026-10-07
author-reported SWE-bench Verified · Write code with agents	80.2	Coding	model author	2026-10-07
author-reported SciCode	52.2	Coding	model author	2026-10-07
author-reported Toolathlon · Call tools and APIs	50	Agentic	model author	2026-10-07
author-reported MathArena/aime_2026	96.4	—	Model Card	2026-04-20
author-reported diamond (Idavidrein/gpqa)	90.5	—	Model Card	2026-04-20
author-reported hle (cais/hle)	34.7	—	Model Card	2026-04-20
author-reported hle (cais/hle)	54	With tools	Model Card	2026-04-20
author-reported MathArena/hmmt_feb_2026	92.7	—	Model Card	2026-04-20
author-reported mmmu_pro_vision (MMMU/MMMU_Pro)	79.4	—	Model Card	2026-04-20
author-reported SWE_Bench_Pro (ScaleAI/SWE-bench_Pro)	58.6	—	Model Card	2026-04-20
author-reported swe_bench_%_resolved (SWE-bench/SWE-bench_Verified)	80.2	—	Model Card	2026-04-20
author-reported terminalbench_2 (harborframework/terminal-bench-2.0)	66.7	—	Model Card	2026-04-20

What the card says

1 cited sentence

Summary generated, cited

The architecture utilizes a Mixture-of-Experts design with 384 experts.^[1]^[2]

is mixture of experts
“Number of Experts | 384”

Generated 2026-10-07 14:59 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

7 use cases

What it is good at 7 use cases

Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.

Facts

23 fields

Listed in LLMs

Facts

Kimi-K2.6 has 1.03T parameters. (safetensors metadata, observed 2026-10-24)
Kimi-K2.6 is a mixture-of-experts model with 384 experts, routing 8 per token. (config.json, observed 2026-10-24)
Context length is 262,144 tokens (256K). (config.json max_position_embeddings, observed 2026-10-24)
License: Kimi (Modified MIT) (modified-mit) (friction: gated-threshold). MIT plus a display-the-model-name attribution requirement for products above 100M MAU or US$20M monthly revenue; the exact wording is in each repo's LICENSE. (license file, observed 2026-10-24)
Created 2026-04-14, last modified 2026-05-19. (HF API, observed 2026-10-24)

Pipeline tag	image-text-to-text
Library	transformers
Architectures	KimiK25ForConditionalGeneration
Model type	kimi_k25
Dtype	bfloat16
Params by dtype	BF16: 12,193,329,648 · F32: 23,040 · I32: 1,014,686,023,680
Size band	XXL
MoE	yes
Experts	384
Experts per token	8
Context length	262,144 (config.json max_position_embeddings)
Papers	arXiv:2602.02276
License	Kimi (Modified MIT) (modified-mit) (other)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · KimiK25ForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · KimiK25ForConditionalGeneration · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for Kimi-K2.6: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	2.46 TB · Cluster (≥640 GB)
FP8 (1 byte/param)	1.23 TB · Cluster (≥640 GB)
INT4 (0.5 bytes/param)	616 GB · Multi-GPU node (<640 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra. MoE: all experts must be resident, so total parameters count.

All variants

40 variants

Lineage

Original release (shipped quantized by the lab) Family: moonshotai/Kimi-K2.6 74 derivatives on the Hub

quantized: 45finetune: 18merge: 2adapter: 9

Top variants (40 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
nvidia/Kimi-K2.6-NVFP4	quantized	NVFP4	—	—	other	nvidia	47	2.3M
unsloth/Kimi-K2.6-GGUF	quantized	GGUF	—	—	other	unsloth	176	985.1k
lightseekorg/kimi-k2.6-eagle3-mla	finetune	—	3.02B	M	other	lightseekorg	7	501.5k
mlx-community/Kimi-K2.6-mlx-DQ3_K_M-q8	quantized	mlx	1.03T	XXL	other	mlx-community	9	369k
amd/Kimi-K2.6-MXFP4	quantized	MXFP4	550B	XXL	other	amd	3	93.7k
unsloth/Kimi-K2.6	quantized	—	1.06T	XXL	other	unsloth	18	84.8k
RedHatAI/Kimi-K2.6-NVFP4	quantized	NVFP4	—	—	other	RedHatAI	13	43.6k
RedHatAI/Kimi-K2.6-FP8-BLOCK	quantized	-FP8	1.03T	XXL	other	RedHatAI	4	43.9k
wafer-ai/Kimi-K2.6-NVFP4	quantized	NVFP4	—	—	other	wafer-ai	3	31.6k
bartowski/moonshotai_Kimi-K2.6-GGUF	quantized	GGUF	—	—	other	bartowski	4	20.8k
nvidia/Kimi-K2.6-Eagle3	finetune	—	1.84B	S	other	nvidia	17	18.6k
ubergarm/Kimi-K2.6-GGUF	quantized	GGUF	—	—	other	ubergarm	38	8.8k
AesSedai/Kimi-K2.6-GGUF	quantized	GGUF	—	—	—	AesSedai	14	9.6k
batiai/Kimi-K2.6-GGUF	quantized	GGUF	—	—	other	batiai	0	8.2k
spicyneuron/Kimi-K2.6-MLX-3.6bit	quantized	MLX	1.03T	XXL	—	spicyneuron	0	7.8k
nvidia/Kimi-K2.6-DFlash	finetune	—	3.48B	M	other	nvidia	27	5.7k
aidiffuser/Kimi-K2.6-MLX-4bit	quantized	MLX	1.03T	XXL	other	aidiffuser	1	6.1k
CosineAI/lumen-outpost-2026-04-27	finetune	—	1.03T	XXL	other	CosineAI	2	5.9k
DevQuasar/moonshotai.Kimi-K2.6-GGUF	quantized	GGUF	—	—	—	DevQuasar	0	6k
festr2/kimi-k2.6-eagle3-mla-fp8	quantized	-fp8	3.02B	M	other	festr2	1	5k
spicyneuron/Kimi-K2.6-MLX-3.3bit	quantized	MLX	1.03T	XXL	—	spicyneuron	0	4.9k
freakyskittle/kimi-k2.75-code-GGUF	merge	GGUF	—	—	other	freakyskittle	1	4.2k
sombra-x/Kimi-K2.6-REAP-Solidity	quantized	—	688B	XXL	other	sombra-x	0	2.8k
novita/kimi-k2.6-dspark	finetune	—	3.15B	M	—	novita	18	1.8k
Deepdive404/Kimi-K2.6-GGUF	quantized	GGUF	—	—	other	Deepdive404	0	2.5k
wafer-ai/Kimi-K2.6-MXFP4	quantized	MXFP4	551B	XXL	other	wafer-ai	2	2.2k
baa-ai/Kimi-K2.6-RAM-344GB-GGUF	quantized	GGUF	—	—	other	baa-ai	1	2.1k
baa-ai/Kimi-K2.6-RAM-GGUF	quantized	GGUF	—	—	other	baa-ai	0	2k
webbrain-one/DeepSeek-V4-Flash-Vision-NVFP4	quantized	NVFP4	291B	XXL	mit	webbrain-one	20	840
Shadiamr/Kimi-K2.6-GGUF-shadoo	quantized	GGUF	—	—	other	Shadiamr	0	1.8k
gghfez/Kimi-K2.6-control-vectors	adapter	GGUF	—	—	cc-by-nc-4.0	gghfez	2	1.6k
freakyskittle/kimi-k2.75-code	merge	—	742B	XXL	other	freakyskittle	1	1.5k
freakyskittle/Kimi-K2.6-DFlash-baseinit	quantized	GGUF	—	—	mit	freakyskittle	0	1.3k
tiny-random/kimi-k2.6	finetune	—	4.47M	S	—	tiny-random	1	1.2k
Panchovix/Kimi-K2.6-GGUF	quantized	GGUF	—	—	—	Panchovix	0	1.2k
Youssofal/Kimi-K2.6-Abliterated-Heretic-GGUF	finetune	GGUF	—	—	other	Youssofal	23	0
Youssofal/Kimi-K2.6-GGUF	quantized	GGUF	—	—	other	Youssofal	1	980
mlx-community/Kimi-K2.6-MoE-Smart-Quant	quantized	-Quant	—	—	—	mlx-community	16	0
florianleibert/kimi-k26-dflash-mi300x	finetune	—	—	—	apache-2.0	florianleibert	15	0
haanjack/Kimi-K2.6-MXFP4	quantized	MXFP4	551B	XXL	apache-2.0	haanjack	0	649

Repository changelog

15 commits

changed in the last 7 daysRank band moves: 2026-10-23Other: 2026-10-24all changes

Repository changelog 15 commits

Date	Commit	SHA
2026-05-19	Add MMMU-Pro evaluation result (#41)	7eb5002
2026-05-12	Revert "use-fast-tokenizer (#38)"	b5aabbf
2026-05-11	use-fast-tokenizer (#38)	81bcaaa
2026-04-30	support-transformers-inference (#26)	2755962
2026-04-29	fix: add missing use_deterministic_attn parameter to MoonViT3dEncoder (#22)	9e6eed7
2026-04-23	Update README.md	5a49d03
2026-04-21	Update SGLang deployment guide: point to K2.6 cookbook and use stable release (#6)	bc8e704
2026-04-20	fix: typo	2b2b88e
2026-04-20	Add evaluation results for HLE, GPQA, AIME, HMMT, SWE-Bench, and Terminal-Bench (#4)	d9cb81b
2026-04-20	add tech blog	36d7872
2026-04-20	update footnote	e69e669
2026-04-20	Update README.md	20008c5
2026-04-20	Update README.md	4931c56
2026-04-14	Add files using upload-large-folder tool	e0ee936
2026-04-14	initial commit	344a634

All providers

18 providers · 18 endpoints

Where to run 18 providers · 18 endpoints

Hosted endpoints observed for Kimi-K2.6, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.45/M.

Provider	Model ref	Context	Quant	$/M in	$/M out	tok/s	TTFT	Tools	Status	Observed
Baidu openrouter	moonshotai/kimi-k2.6	256K	fp4	$0.45	$1.89	—	—	yes	live	2026-10-24
Chutes openrouter	moonshotai/kimi-k2.6	256K	int4	$0.45	$1.95	—	—	yes	live	2026-10-24
DigitalOcean openrouter	moonshotai/kimi-k2.6	256K	unknown	$0.57	$2.40	—	—	yes	live	2026-10-24
StreamLake openrouter	moonshotai/kimi-k2.6	250K	fp8	$0.60	$2.52	—	—	yes	live	2026-10-24
Inceptron openrouter	moonshotai/kimi-k2.6	256K	int4	$0.60	$3.20	—	—	yes	live	2026-10-24
CoreWeave openrouter	moonshotai/kimi-k2.6	256K	fp4	$0.65	$3.41	—	—	yes	live	2026-10-24
Crusoe openrouter	moonshotai/kimi-k2.6	256K	bf16	$0.70	$3.50	—	—	yes	live	2026-10-24
DeepInfra hf, deepinfra, openrouter	moonshotai/Kimi-K2.6	256K	fp4	$0.75	$3.50	38.1	414 ms	yes	live	2026-10-25
Parasail parasail, openrouter	parasail-kimi-k26	256K	int4	$0.75	$3.50	—	—	yes	serverless	2026-10-24
Venice openrouter	moonshotai/kimi-k2.6	250K	int4	$0.75	$3.50	—	—	yes	live	2026-10-24
SiliconFlow openrouter	moonshotai/kimi-k2.6	256K	fp8	$0.77	$3.40	—	—	yes	live	2026-10-24
Novita hf, openrouter	moonshotai/kimi-k2.6	256K	unknown	$0.80	$3.40	40.9	2050 ms	yes	live	2026-10-25
GMICloud openrouter	moonshotai/kimi-k2.6	256K	fp8	$0.85	$3.60	—	—	yes	live	2026-10-24
AtlasCloud openrouter	moonshotai/kimi-k2.6	256K	int4	$0.95	$4.00	—	—	yes	live	2026-10-24
Cloudflare openrouter	moonshotai/kimi-k2.6	256K	unknown	$0.95	$4.00	—	—	yes	live	2026-10-24
Moonshot AI openrouter	moonshotai/kimi-k2.6	256K	int4	$0.95	$4.00	—	—	yes	live	2026-10-24
Phala openrouter	moonshotai/kimi-k2.6	256K	unknown	$1.09	$4.60	—	—	yes	live	2026-10-24
Featherless hf	moonshotai/Kimi-K2.6	—	—	—	—	—	—	—	live	2026-10-25

Articles and coverage

24 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Official4

Kimi K2.6: Advancing open-source coding

kimi.com2026-04-20officialauthorityHN front pageHN 710 pts · 372 comments2,299 words

Kimi K2.6 is an open-source model released for advanced coding, long-horizon execution, and agent swarm capabilities. It is available via Kimi.com, the Kimi App, the API, and Kimi Code, showing a 12% increase in code generation accuracy and a 96.60% tool invocation success rate. The model scales agent swarms to 300 sub-agents executing across 4,000 coordinated steps simultaneously.written, reviewed, cited

Kimi K2.6 optimizes model inference in Zig, a highly niche programming language, to demonstrate exceptional out-of-distribution generalization.
The model reconfigured the exchange-core thread topology from 4ME+2RE to 2ME+1RE to extract a 185% medium throughput leap.

Kimi K2.6-code-preview is now available

kimi.com2026-04-13officialauthorityHN 14 pts · 5 comments165 words

Kimi Code is an AI coding assistant built by Moonshot AI that runs on a user's computer. The tool can write and edit code, understand codebases, run commands, and look things up.generated, cited

Kimi K2.6 Tech Blog: Advancing Open-Source Coding

kimi.comofficialauthority2,299 words

Kimi K2.6 optimized Zig inference to achieve speeds ~20% faster than LM Studio.
The model reconfigured an 8-year-old engine's thread topology from 4ME+2RE to 2ME+1RE.

Overview of Kimi K2.6 Model

platform.kimi.ai2026-04-21officialHN 2 pts · 0 comments808 words

Analysis5

Kimi K2.6 just beat Claude, GPT-5.5, and Gemini in a coding challenge

thinkpol.ca2026-05-03HN front pageHN 380 pts · 219 comments1,401 words

Kimi K2.6 won the Word Gem Puzzle challenge with 22 match points and a 7-1-0 record, beating GPT-5.5 and Claude Opus 4.7. The model used a greedy approach to slide tiles on grids ranging from 10×10 to 30×30, outperforming Western frontier labs. Kimi K2.6 is open-weights and publicly available from Moonshot AI.written, reviewed, cited

Nvidia's Nemotron Super 3 produced code with a syntax error that prevented connection to the game server.
Muse Spark claimed every visible word regardless of length, resulting in a cumulative score of −15,309.
MiMo V2-Pro never slid a single tile because its threshold for positive value was never triggered.

US lawmakers investigate DoorDash's use of Moonshot AI's Kimi K2.6 model

scmp.com2026-08-01HN 11 pts · 4 comments

Kimi K2.6 Intelligence, Performance and Price Analysis

artificialanalysis.ai2026-04-21authorityHN 3 pts · 0 comments1,066 words

Berget AI announces Berget Code for European teams powered by Kimi K2.6

berget.ai2026-05-13HN 5 pts · 3 comments459 words

Claude-Code-Proxy – Use Kimi K2.6 and OpenAI Subscriptions in Claude Code

github.com2026-05-04authorityHN 2 pts · 0 comments446 words

Community1

Kimi K2.6: Advancing Open-Source Coding

twitter.com2026-04-20HN 40 pts · 1 comments

All coverage tracked for this model (24) — append-only, nothing is ever removed

Title	Source	Published	First seen	Signals
Kimi K2.6 Tech Blog: Advancing Open-Source Coding	kimi.com official	—	2026-10-07	exa
Kimi-K2.6: release, architecture & GPU deployment \| AIInfraInterviews	aiinfrainterviews.com	—	2026-10-07	exa
Kimi K2.6 — AI Pro Playbook \| AI Pro Playbook	aiproplaybook.com	—	2026-10-07	exa
US lawmakers investigate DoorDash's use of Moonshot AI's Kimi K2.6 model	scmp.com	2026-08-01	2026-10-07	hn
Kimi K2.6 Deep Dive: What Makes This Model Different?	aiindigo.com	2026-06-19	2026-10-07	exa
Kimi K2.6 Deep Dive: Is it the New King of Open Weights?	aiindigo.com	2026-06-14	2026-10-07	exa
Kimi K2.6, reviewed — benchr	benchr.org	2026-05-30	2026-10-07	exa
Cerebras Brings Trillion Parameter Inference to Enterprises with Kimi K2.6	cerebras.ai	2026-05-20	2026-10-07	hn
DeepSeek V4 Pro and Flash vs. Claude Opus 4.7 and Kimi K2.6	blog.kilo.ai	2026-05-15	2026-10-07	hn
Berget AI announces Berget Code for European teams powered by Kimi K2.6	berget.ai	2026-05-13	2026-10-07	hn
Kimi K2.6 Review: Moonshot AI's Open-Weight Trillion-Parameter Agent — Agent Swarm, MLA Attention, 80.2% SWE-Bench — ChatForest	chatforest.com	2026-05-13	2026-10-07	exa
Show HN: Kimiflare, Kimi K2.6 as a Coding Agent on Cloudflare Workers AI	news.ycombinator.com	2026-05-12	2026-10-07	hn
Claude-Code-Proxy – Use Kimi K2.6 and OpenAI Subscriptions in Claude Code	github.com	2026-05-04	2026-10-07	hn
Kimi K2.6: Open-Source Just Beat GPT-5.5 at Coding	buildfastwithai.com	2026-05-04	2026-10-07	exa
Kimi K2.6 just beat Claude, GPT-5.5, and Gemini in a coding challenge	thinkpol.ca	2026-05-03	2026-10-07	hn
Kimi K2.6 Review (8.5/10): 1T Open MoE Beats Opus on Cost \| ThePlanetTools.ai	theplanettools.ai	2026-04-29	2026-10-07	exa
We Gave Claude Opus 4.7 and Kimi K2.6 the Same Workflow Orchestration Spec	blog.kilo.ai	2026-04-22	2026-10-07	hn
Kimi K2.6 Intelligence, Performance and Price Analysis	artificialanalysis.ai	2026-04-21	2026-10-07	hn
Kimi K2.6 with Strix: a quick test	theartificialq.github.io	2026-04-21	2026-10-07	hn
Overview of Kimi K2.6 Model	platform.kimi.ai official	2026-04-21	2026-10-07	hn
Kimi K2.6: Advancing Open-Source Coding	twitter.com	2026-04-20	2026-10-07	hn
Kimi K2.6: Advancing open-source coding	kimi.com official	2026-04-20	2026-10-07	hn
Kimi K2.6 Developer Guide: Benchmarks, API & Agent Swarm \| Lushbinary	lushbinary.com	2026-04-20	2026-10-07	exa
Kimi K2.6-code-preview is now available	kimi.com official	2026-04-13	2026-10-07	hn

Papers

1 paper

Papers 1

arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).

Kimi K2.5: Visual Agentic Intelligence this model

Kimi Team, Tongtong Bai, Yifan Bai +334 · 2026-02-02 · PDF · HF paper page (278 upvotes)

Similar models

3 models

Similar models

Inkling · 952B
same size band (XXL), also MoE, same task (image-text-to-text), fewer providers, permissive license
DeepSeek-V4.1-Flash · 763B
same size band (XXL), also MoE, same task (image-text-to-text), more providers, permissive license
Kimi-K2.5 · 1.03T
same size band (XXL), same task (image-text-to-text), fewer providers

Sources

5 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:32 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 14:59 UTC · Last verified 2026-10-25

Follow Kimi-K2.6

Watch this model

/feeds/models/moonshotai--Kimi-K2.6.xml(RSS 2.0) · every change to Kimi-K2.6
/feeds/families/moonshotai--Kimi-K2.6.xml(RSS 2.0) · every change across the Kimi-K2.6 family

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.