Moonshot AIKimi K2KimiK25ForConditionalGeneration

Kimi K2.5 — Hardware Requirements & GPU Compatibility

VisionFunctions

Kimi K2.5 is an earlier release in Moonshot AI's Kimi K2 series, sharing the roughly 1-trillion-parameter scale of later checkpoints in the line. Released at the start of 2026, it accepts image input alongside text, giving it multimodal capability on top of standard chat use. Its context window reaches 256K tokens, enough for lengthy documents or long conversation histories, and it is released under a custom license outside the standard open-source terms. Because of its size, Kimi K2.5 needs multi-GPU or server-class hardware; most people will use it through a hosted endpoint rather than running it themselves.

1026.9B params · MoE33B active262K contextOtherJan 2026394.2K downloads2.9K likes321.5K quant downloads

Your hardware

Verdicts below are for the reference RTX 4090. Set yours to personalise the page.

Detection reads your browser's GPU name with no data sent anywhere.

Kimi K2.5 runs on a multi-GPU server

Q4_K_M needs ~617.6 GB · about 5× H200 SXM 141 GB · or Mac Studio (M3 Ultra, 512GB) at Q3_K_M (~25 tok/s)

Macs from 512 GB (Q3_K_M · ~25 tok/s)

Score 113.9#25 of 106 open#21 of 102 that fit ≤640 GBOther

Not set up for it?

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

llmrun Composite Score →

113.9

90% range 111.2–116.5
#25 of 106 open models
#21 of 102 that fit ≤640 GB

Benchmarks · tier chart ↓

Among 106 open modelsHow the Composite Score is built

Coding110.490% range 103–118
Agents110.490% range 104–117
MathNot enough data yet
Science116.790% range 107–126
Reasoning114.790% range 111–118

Run Kimi K2.5

One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.

Ways to runOther ways

each opens step-by-step

Run in cloud

Rent 5× H200 SXM · Q4_K_M

Estimated hourly price: ~$17.95/hr; est., whole server
GPUs: 5 GPUs; 705 GB of VRAM

Deployon RunPodRent H200 SXM$3.59/hr per GPU

$5+ bonus for new users

Compare GPUs →

How Much VRAM Does Kimi K2.5 Need?

Select a quantization level; the listed hardware adheres to it.

VRAM	+32K	+128K	File	Quality	Runs on
286.3 GB	285 GB	291.9 GB	326.7 GB	2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss	Runs on Instinct MI350X 288 GB
311.9 GB	310.7 GB	317.6 GB	283.1 GB	2-bit · i-quant extra small — Importance-weighted 2-bit, extra small	Runs on 5× H100 SXM 80 GB
324.8 GB	323.5 GB	330.4 GB	283.8 GB	2-bit · i-quant small — Importance-weighted 2-bit, small	Runs on 5× H100 SXM 80 GB
350.4 GB	349.2 GB	356.1 GB	345.2 GB	2-bit · i-quant medium — Importance-weighted 2-bit, medium	Runs on 5× H100 SXM 80 GB
401.8 GB	400.5 GB	407.4 GB	414.8 GB	3-bit · i-quant — Importance-weighted 3-bit	Runs on 6× H100 SXM 80 GB
427.5 GB	426.2 GB	433.1 GB	420.1 GB	3-bit · i-quant extra small — Importance-weighted 3-bit, extra small	Runs on 6× H100 SXM 80 GB
440.3 GB	439 GB	445.9 GB	405.3 GB	3-bit · i-quant small — Importance-weighted 3-bit, small	Runs on 7× H100 SXM 80 GB
440.3 GB	439 GB	445.9 GB	373.8 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Runs on 7× H100 SXM 80 GB
453.1 GB	451.9 GB	458.8 GB	443.2 GB	3-bit small — 3-bit small quantization	Runs on 7× H100 SXM 80 GB
466.0 GB	464.7 GB	471.6 GB	466.9 GB	3-bit · i-quant medium — Importance-weighted 3-bit, medium	Runs on 7× H100 SXM 80 GB
504.5 GB	503.2 GB	510.1 GB	489.8 GB	3-bit medium — 3-bit medium quantization	Runs on 7× H100 SXM 80 GB
517.3 GB	516 GB	523 GB	581.2 GB	4-bit · legacy — 4-bit legacy quantization	Runs on 8× H100 SXM 80 GB
530.2 GB	528.9 GB	535.8 GB	487.5 GB	3-bit large — 3-bit large quantization	Runs on 8× H100 SXM 80 GB
555.8 GB	554.6 GB	561.5 GB	547.2 GB	4-bit · i-quant compact — Importance-weighted 4-bit, compact	Runs on 8× H100 SXM 80 GB
581.5 GB	580.2 GB	587.1 GB	579.1 GB	4-bit · i-quant non-linear — Importance-weighted 4-bit, non-linear	Runs on 5× H200 SXM 141 GB
581.5 GB	580.2 GB	587.1 GB	642.9 GB	4-bit · legacy+ — 4-bit legacy quantization with offset	Runs on 5× H200 SXM 141 GB
581.5 GB	580.2 GB	587.1 GB	583.3 GB	4-bit small — 4-bit small quantization	Runs on 5× H200 SXM 141 GB
620.0 GB	618.7 GB	625.6 GB	621.2 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Runs on 5× H200 SXM 141 GB
632.9 GB	631.6 GB	638.5 GB	581.2 GB	4-bit large — 4-bit large quantization	Runs on 5× H200 SXM 141 GB
709.9 GB	708.6 GB	715.5 GB	707 GB	5-bit small — 5-bit small quantization	Runs on 6× H200 SXM 141 GB
735.5 GB	734.3 GB	741.2 GB	728.7 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Runs on 6× H200 SXM 141 GB
851.1 GB	849.8 GB	856.7 GB	842.9 GB	6-bit, very good — 6-bit quantization, very good quality	Runs on 7× H200 SXM 141 GB
1030.8 GB	1029.5 GB	1036.4 GB	1091.1 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Runs on 6× B200 192 GB
2057.6 GB	2056.4 GB	2063.3 GB	2053.2 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Runs on 4 nodes of 8× H100 SXM

Selected Q4_K_M

4-bit medium quantization

617.6GBat 16K context

Weights: 616.1 GB
KV cache: 1.2 GB
Overhead: 0.3 GB

Server-class: ~617.6 GB at 16K context · about 5× H200 SXM 141 GB

Which Hardware Can Run Kimi K2.5?

Q4_K_M · 620.0 GB

Kimi K2.5 (Q4_K_M) requires 620.0 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 807+ GB is recommended. Using the full 262K context window can add up to 14.8 GB, bringing total usage to 634.9 GB. No consumer GPU has enough memory.

Needs server-class hardware

Kimi K2.5 runs on a multi-GPU server · Q4_K_M needs ~617.6 GB

Hardware that runs this model
Hardware	Quant	Memory	Speed	Context	Hardware page
Mac 512 GB · Mac Studio (M3 Ultra, 512GB)	Q3_K_M	502.1 GB	~25 tok/s	64K ctx	Mac Studio →	Q3_K_M · 502.1 GB · ~25 tok/s · 64K ctx
8× H100 SXM 80 GB · 640 GB	IQ4_XS	553.4 GB	—	256K ctx	H100 SXM →	IQ4_XS · 553.4 GB · 256K ctx
5× H200 SXM 141 GB · 705 GB	Q4_K_M	617.6 GB	—	128K ctx	H200 SXM →	Q4_K_M · 617.6 GB · 128K ctx
4× B200 192 GB · 768 GB	Q4_K_M	617.6 GB	—	256K ctx	B200 →	Q4_K_M · 617.6 GB · 256K ctx

Also: NVIDIA DGX A100 640GB · NVIDIA DGX H100

≤ 640 GB tier →

By VRAM tier · typical GPUs and Macs at Q4_K_M

Consumer GPUs (up to 32 GB) and Macs up to 256 GB: too small for this model (needs ≥ 283.8 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →

How good is Kimi K2.5?

The llmrun Score merges 19 public benchmark boards into a single figure, where 100 represents the average for open models.

Score vs VRAM in the ≤640 GB tier

Kimi K2.5
Other models
Best Score for its VRAM
x: VRAM (GB, log scale) · y: llmrun Score

#21 of 102 that fit ≤640 GB

Each dot is a model at the best quant that fits ≤640 GB. The line steps up to the best Score available at each amount of VRAM; a dot below it is beaten by a model that needs no more memory.

Full ≤640 GB ranking →

Benchmarks

SWE-bench Verified
SWE-bench · Feb 2026Independent
70.8%
#4/20 open
Independent
GPQA Diamond
Epoch AI · Jan 2026Independent
87.6%
#19/100 open
Independent
LMArena Text
LMArena · Oct 2026Arena
1449.7
#15/188 open
Arena
AIME 2024/2025
Epoch AI · Jan 2026Independent
92.2%
#13/88 open
Independent
Terminal-Bench
Epoch AI · Jan 2026Lab-reported
43.2%
#3/16 open
Lab-reported
Humanity's Last Exam
Epoch AI · Jan 2026Lab-reported
24.4%
#1/4 open
Lab-reported

All 21 Kimi K2.5 benchmarks

In the ≤640 GB class

102 models fit ≤640 GB

Same Score, only models that fit ≤640 GB at their best quant: this model's peers by memory need.

#1
DeepSeek V4.1 Flash 1017DeepSeek
134.4
458.4 GB
#2
MiMo V2.6 Pro RLXiaomi
133.3
614.1 GB
#3
MiMo V2.6 Flash RLXiaomi
131.9
186.5 GB
#4
GLM 5.3Z.ai
130.1
453.8 GB
#5
DeepSeek V4.1 FlashDeepSeek
129.4
458.4 GB
#6
Step 5 PreviewStepFun
128.0
407.5 GB

Full ≤640 GB ranking →

Alternatives to Kimi K2.5

Kimi K2 family · 8 models →

Official models with an llmrun Score, each answering one question; sizes are each model's default quant (usually Q4_K_M) with a 16K-token context.

Best alternative at this size
MiMo V2.6 Pro RL
Xiaomi
- 19.4 higher Score
- 3.5 GB less VRAM
Score 133.3 · 614.1 GB
Compare with Kimi K2.5 Open MiMo V2.6 Pro RL
Fits in less VRAM
DeepSeek V4.1 Flash 1017
DeepSeek
- 20.5 higher Score
- 159.2 GB less VRAM
Score 134.4 · 458.4 GB
Compare with Kimi K2.5 Open DeepSeek V4.1 Flash 1017
Other vendor
GLM 5.3
Z.ai
- 16.2 higher Score
- 163.8 GB less VRAM
Score 130.1 · 453.8 GB
Compare with Kimi K2.5 Open GLM 5.3

Similar models

Model	Score	VRAM	tok/s	Why it is similar
DeepSeek V4.1 FlashDeepSeek	129.4	458.4 GB	—needs > 24 GB	similar score
DeepSeek V4 Pro 0813DeepSeek	129.2	990.8 GB	—needs > 24 GB	similar score
Step 5 PreviewStepFun	128.0	407.5 GB	—needs > 24 GB	similar score
Muse Spark Open 816B A42BMeta	126.8	551.9 GB	—needs > 24 GB	similar score
GLM 5.2Z.ai	122.9	453.8 GB	—needs > 24 GB	similar score
Kimi K2.6Moonshot AI	120.9	617.6 GB	—needs > 24 GB	similar score
GLM 5.1Z.ai	117.8	454.1 GB	—needs > 24 GB	similar score
DeepSeek V4 ProDeepSeek	116.7	959.8 GB	—needs > 24 GB	similar score
DeepSeek V4 FlashDeepSeek	115.3	175.1 GB	—needs > 24 GB	similar score
MiniMax M3MiniMax	112.6	258 GB	—needs > 24 GB	similar score

Specifications

Publisher: Moonshot AI
Family: Kimi K2
Parameters: 1026.9B
Active parameters: 33.3B per token
Architecture: KimiK25ForConditionalGeneration
Context Length: 262,144 tokens
Vocabulary Size: 163,840
Release Date: 2026-01-01
License: Other

Resources

Frequently Asked Questions

How much VRAM does Kimi K2.5 need?

Kimi K2.5 requires 620.0 GB of VRAM at Q4_K_M, or 2057.6 GB at BF16. Full 262K context adds up to 14.8 GB (634.9 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 1026.9B × 4.8 bits ÷ 8 = 616.1 GB

KV Cache + Overhead ≈ 3.9 GB (at 2K context + ~0.3 GB framework)

Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.

KV Cache + Overhead ≈ 18.8 GB (at full 262K context)

VRAM usage by quantization

Q4_K_M

620.0 GB

Q4_K_M + full context

634.9 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 5090 run Kimi K2.5?

No — Kimi K2.5 requires at least 286.3 GB at IQ2_XXS, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.

What's the best quantization for Kimi K2.5?

For Kimi K2.5, Q4_K_M (620.0 GB) offers the best balance of quality and VRAM usage. Q4_K_L (632.9 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 286.3 GB.

VRAM requirement by quantization

IQ2_XXS

286.3 GB~53%

IQ3_S

440.3 GB~75%

Q3_K_L

530.2 GB~86%

Q4_K_M ★

620.0 GB~89%

Q4_K_L

632.9 GB~90%

BF16

2057.6 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Kimi K2.5 on a Mac?

Yes, but only at lower quantizations. The smallest Mac that can run Kimi K2.5 is Mac Studio (M3 Ultra, 512GB) at Q3_K_M; 1 of the 146 Macs we list can run it at some quantization. For Q4_K_M (620.0 GB) you need a Mac with more unified memory.

Can I run Kimi K2.5 locally?

Yes — Kimi K2.5 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

What's the download size of Kimi K2.5?

At Q4_K_M, the download is about 616.13 GB. The full-precision BF16 version is 2053.76 GB. The smallest option (IQ2_XXS) is 282.39 GB.

Which GPUs can run Kimi K2.5?

No single consumer GPU has enough VRAM to run Kimi K2.5 at Q4_K_M (620.0 GB). Multi-GPU or professional hardware is required.

Which devices can run Kimi K2.5?

Kimi K2.5 requires at least 620.0 GB at Q4_K_M, which exceeds the usable memory of every unified-memory device we list (the largest has 508.5 GB).