Multi-model provider comparison

MiniMax M3, Kimi K2.6 & GLM 5.2 Provider Comparison

This model set is designed for teams evaluating several current open-weight text models behind one provider relationship. The comparison requires complete coverage before a provider can win or receive a metric rank.

At a glance

Complete-provider winners

Every winner offers all selected models. The recommendation uses the single provider closest to the current ideal combination of lower total cost and shorter average response time.

Cost/speed balance

DeepInfra3/3 selected models

12.62 s average response

$0.00583 per model-set run · 1K input + 500 output per model

Cheapest complete provider

DeepInfra3/3 selected models

$1.2956 / 1M tokens

Blended 1K input + 500 output mix, averaged across 3 selected models

Fastest complete provider

Venice3/3 selected models

9.10 s

Average estimated 500-token response across 3 models

Rank complete providers by

Provider Speed measurements Reliability & location

Offers all 3 models · 6 providers

Provider

DeepInfraRecommended

Offers all 3 models

Speed measurements

Average 500-token response: 12.62 s
Average response start: 0.92 s
Average output speed: 53.3 tok/s

Reliability & location

Lowest route uptime: 98.82%
Headquarters: United States
Regions: N/a

Review DeepInfra routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	`deepinfra/bf16`	Input $0.3 / 1M tokens Output $1.2 / 1M tokens	12.89 s	1.26 s	43.0 tok/s	99.70%	524.3K
Kimi K2.6moonshotai/kimi-k2.6	`deepinfra/fp4`	Input $0.75 / 1M tokens Output $3.5 / 1M tokens	6.34 s	0.72 s	89.0 tok/s	99.83%	262.1K
GLM 5.2z-ai/glm-5.2	`deepinfra/fp4`	Input $0.93 / 1M tokens Output $3 / 1M tokens	18.64 s	0.78 s	28.0 tok/s	98.82%	1M

Provider

AtlasCloud

Offers all 3 models

Speed measurements

Average 500-token response: 24.83 s
Average response start: 1.62 s
Average output speed: 21.7 tok/s

Reliability & location

Lowest route uptime: 96.17%
Headquarters: United States
Regions: N/a

Review AtlasCloud routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	`atlas-cloud/fp8`	Input $0.3 / 1M tokens Output $1.2 / 1M tokens	24.78 s	0.97 s	21.0 tok/s	96.17%	524.3K
Kimi K2.6moonshotai/kimi-k2.6	`atlas-cloud/int4`	Input $0.95 / 1M tokens Output $4 / 1M tokens	22.81 s	1.98 s	24.0 tok/s	97.34%	262.1K
GLM 5.2z-ai/glm-5.2	`atlas-cloud/fp8`	Input $1.26 / 1M tokens Output $3.96 / 1M tokens	26.91 s	1.91 s	20.0 tok/s	99.69%	1M

Provider

NovitaAI

Offers all 3 models

Speed measurements

Average 500-token response: 18.84 s
Average response start: 2.62 s
Average output speed: 33.7 tok/s

Reliability & location

Lowest route uptime: 97.04%
Headquarters: United States
Regions: N/a

Review NovitaAI routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	`novita/fp8`	Input $0.3 / 1M tokens Output $1.2 / 1M tokens	12.37 s	2.17 s	49.0 tok/s	99.40%	1M
Kimi K2.6moonshotai/kimi-k2.6	`novita`	Input $0.8 / 1M tokens Output $3.4 / 1M tokens	21.48 s	2.25 s	26.0 tok/s	99.62%	262.1K
GLM 5.2z-ai/glm-5.2	`novita/fp8`	Input $0.9744 / 1M tokens Output $3.06 / 1M tokens	22.65 s	3.42 s	26.0 tok/s	97.04%	1M

Provider

Parasail

Offers all 3 models

Speed measurements

Average 500-token response: 13.53 s
Average response start: 0.76 s
Average output speed: 45.3 tok/s

Reliability & location

Lowest route uptime: 93.16%
Headquarters: United States
Regions: N/a

Review Parasail routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	`parasail/fp8`	Input $0.3 / 1M tokens Output $1.2 / 1M tokens	20.07 s	0.84 s	26.0 tok/s	93.16%	1M
Kimi K2.6moonshotai/kimi-k2.6	`parasail/int4`	Input $0.75 / 1M tokens Output $3.5 / 1M tokens	8.21 s	0.75 s	67.0 tok/s	99.44%	262.1K
GLM 5.2z-ai/glm-5.2	`parasail/fp4`	Input $1.4 / 1M tokens Output $4.4 / 1M tokens	12.31 s	0.68 s	43.0 tok/s	100.00%	262.1K

Provider

Together

Offers all 3 models

Speed measurements

Average 500-token response: 10.42 s
Average response start: 0.65 s
Average output speed: 64.0 tok/s

Reliability & location

Lowest route uptime: 99.02%
Headquarters: United States
Regions: N/a

Review Together routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	`together`	Input $0.3 / 1M tokens Output $1.2 / 1M tokens	16.83 s	0.70 s	31.0 tok/s	99.02%	524.3K
Kimi K2.6moonshotai/kimi-k2.6	`together`	Input $1.2 / 1M tokens Output $4.5 / 1M tokens	5.55 s	0.55 s	100.0 tok/s	99.73%	262.1K
GLM 5.2z-ai/glm-5.2	`together`	Input $1.4 / 1M tokens Output $4.4 / 1M tokens	8.89 s	0.70 s	61.0 tok/s	99.51%	262.1K

Provider

Venice

Offers all 3 models

Speed measurements

Average 500-token response: 9.10 s
Average response start: 1.21 s
Average output speed: 73.3 tok/s

Reliability & location

Lowest route uptime: Insufficient recent data
Headquarters: United States
Regions: United States

Review Venice routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	`venice/fp8`	Input $0.3 / 1M tokens Output $1.2 / 1M tokens	6.10 s	1.29 s	104.0 tok/s	96.45%	524.3K
Kimi K2.6moonshotai/kimi-k2.6	`venice/int4`	Input $0.75 / 1M tokens Output $3.5 / 1M tokens	7.68 s	1.02 s	75.0 tok/s	Insufficient recent data	256K
GLM 5.2z-ai/glm-5.2	`venice/fp8`	Input $1.4 / 1M tokens Output $4.4 / 1M tokens	13.51 s	1.31 s	41.0 tok/s	98.71%	1M

Missing selected models · 24 providers

Provider

Baidu Qianfan

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: China
Regions: N/a

Review Baidu Qianfan routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`baidu/fp4`	Input $0.684 / 1M tokens Output $3.42 / 1M tokens	14.90 s	1.01 s	36.0 tok/s	99.94%	262.1K
GLM 5.2z-ai/glm-5.2	`baidu/fp8`	Input $0.97 / 1M tokens Output $3.07 / 1M tokens	11.82 s	1.18 s	47.0 tok/s	99.52%	1M

Missing from Baidu Qianfan: MiniMax M3.

Provider

Baseten

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: N/a

Review Baseten routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`baseten/fp4`	Input $0.95 / 1M tokens Output $4 / 1M tokens	5.24 s	1.07 s	120.0 tok/s	Insufficient recent data	262K
GLM 5.2z-ai/glm-5.2	`baseten/fp8`	Input $1.4 / 1M tokens Output $4.4 / 1M tokens	7.72 s	0.97 s	74.0 tok/s	99.93%	256K

Missing from Baseten: MiniMax M3.

Provider

Cloudflare

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: N/a

Review Cloudflare routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`cloudflare`	Input $0.95 / 1M tokens Output $4 / 1M tokens	11.70 s	0.59 s	45.0 tok/s	99.43%	262.1K
GLM 5.2z-ai/glm-5.2	`cloudflare`	Input $1.4 / 1M tokens Output $4.4 / 1M tokens	9.47 s	0.70 s	57.0 tok/s	99.97%	262.1K

Missing from Cloudflare: MiniMax M3.

Provider

Decart

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: United States

Review Decart routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`decart/fp4`	Input $0.66 / 1M tokens Output $3.41 / 1M tokens	15.21 s	5.78 s	53.0 tok/s	98.13%	262.1K
GLM 5.2z-ai/glm-5.2	`decart/fp4`	Input $1.2 / 1M tokens Output $4.2 / 1M tokens	7.76 s	0.61 s	70.0 tok/s	96.20%	1M

Missing from Decart: MiniMax M3.

Provider

DigitalOcean

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: N/a
Regions: N/a

Review DigitalOcean routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`digitalocean`	Input $0.76 / 1M tokens Output $3.2 / 1M tokens	10.50 s	0.69 s	51.0 tok/s	100.00%	262.1K
GLM 5.2z-ai/glm-5.2	`digitalocean`	Input $1.05 / 1M tokens Output $4.4 / 1M tokens	12.35 s	1.24 s	45.0 tok/s	97.62%	262.1K

Missing from DigitalOcean: MiniMax M3.

Provider

Fireworks

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: N/a

Review Fireworks routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`fireworks`	Input $0.95 / 1M tokens Output $4 / 1M tokens	11.18 s	0.77 s	48.0 tok/s	Insufficient recent data	262.1K
GLM 5.2z-ai/glm-5.2	`fireworks`Alternatives: fireworks/fast	Input $1.4 / 1M tokens Output $4.4 / 1M tokens	11.10 s	1.10 s	50.0 tok/s	99.22%	1M

Missing from Fireworks: MiniMax M3.

Provider

GMICloud

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: United States

Review GMICloud routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	`gmicloud/fp8`	Input $0.3 / 1M tokens Output $1.2 / 1M tokens	6.33 s	1.48 s	103.0 tok/s	99.91%	1M
Kimi K2.6moonshotai/kimi-k2.6	Not offered	—	—	—	—	—	—
GLM 5.2z-ai/glm-5.2	`gmicloud/fp8`	Input $0.98 / 1M tokens Output $3.08 / 1M tokens	15.09 s	1.58 s	37.0 tok/s	97.84%	1M

Missing from GMICloud: Kimi K2.6.

Provider

Inceptron

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: Sweden
Regions: Finland

Review Inceptron routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`inceptron/int4`	Input $0.66 / 1M tokens Output $3.41 / 1M tokens	30.26 s	2.48 s	18.0 tok/s	99.24%	262.1K
GLM 5.2z-ai/glm-5.2	`inceptron/fp4`	Input $0.94 / 1M tokens Output $2.9 / 1M tokens	43.27 s	1.61 s	12.0 tok/s	87.29%	1M

Missing from Inceptron: MiniMax M3.

Provider

Morph

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: United States

Review Morph routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	`morph`	Input $0.6 / 1M tokens Output $2.4 / 1M tokens	35.14 s	1.81 s	15.0 tok/s	95.32%	256K
Kimi K2.6moonshotai/kimi-k2.6	Not offered	—	—	—	—	—	—
GLM 5.2z-ai/glm-5.2	`morph`	Input $1.1 / 1M tokens Output $4.1 / 1M tokens	35.16 s	1.83 s	15.0 tok/s	96.56%	1M

Missing from Morph: Kimi K2.6.

Provider

Phala

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: N/a

Review Phala routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`phala`	Input $1.09 / 1M tokens Output $4.6 / 1M tokens	27.94 s	1.62 s	19.0 tok/s	Insufficient recent data	262.1K
GLM 5.2z-ai/glm-5.2	`phala`	Input $1.4 / 1M tokens Output $4.4 / 1M tokens	38.09 s	2.37 s	14.0 tok/s	37.01%	1M

Missing from Phala: MiniMax M3.

Provider

SiliconFlow

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: Singapore
Regions: United States

Review SiliconFlow routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`siliconflow/fp8`	Input $0.77 / 1M tokens Output $3.4 / 1M tokens	15.29 s	1.78 s	37.0 tok/s	100.00%	262.1K
GLM 5.2z-ai/glm-5.2	`siliconflow/fp8`	Input $1.3 / 1M tokens Output $4.09 / 1M tokens	28.65 s	6.91 s	23.0 tok/s	94.53%	1M

Missing from SiliconFlow: MiniMax M3.

Provider

StreamLake

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: China
Regions: N/a

Review StreamLake routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`streamlake/fp8`	Input $0.855 / 1M tokens Output $3.6 / 1M tokens	19.83 s	1.98 s	28.0 tok/s	97.75%	256K
GLM 5.2z-ai/glm-5.2	`streamlake/fp8`	Input $0.9737 / 1M tokens Output $3.06 / 1M tokens	18.81 s	2.14 s	30.0 tok/s	99.90%	1M

Missing from StreamLake: MiniMax M3.

Provider

Weights & Biases

Offers 2 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: United States

Review Weights & Biases routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`wandb/fp4`	Input $0.95 / 1M tokens Output $4 / 1M tokens	4.33 s	0.33 s	125.0 tok/s	99.95%	262.1K
GLM 5.2z-ai/glm-5.2	`wandb/fp4`	Input $1.39 / 1M tokens Output $4.4 / 1M tokens	5.04 s	0.50 s	110.0 tok/s	99.98%	262.1K

Missing from Weights & Biases: MiniMax M3.

Provider

AkashML

Offers 1 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: N/a
Regions: N/a

Review AkashML routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	Not offered	—	—	—	—	—	—
GLM 5.2z-ai/glm-5.2	`akashml/fp8`	Input $1.3 / 1M tokens Output $4.4 / 1M tokens	17.13 s	1.50 s	32.0 tok/s	99.42%	131.1K

Missing from AkashML: MiniMax M3, Kimi K2.6.

Provider

Alibaba Cloud Int.

Offers 1 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: Singapore
Regions: SingaporeChina

Review Alibaba Cloud Int. routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	Not offered	—	—	—	—	—	—
GLM 5.2z-ai/glm-5.2	`alibaba`	Input $1.24 / 1M tokens Output $3.89 / 1M tokens	18.45 s	1.78 s	30.0 tok/s	99.98%	1M

Missing from Alibaba Cloud Int.: MiniMax M3, Kimi K2.6.

Provider

Ambient

Offers 1 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: N/a
Regions: N/a

Review Ambient routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	Not offered	—	—	—	—	—	—
GLM 5.2z-ai/glm-5.2	`ambient/fp8`	Input $1.05 / 1M tokens Output $4.4 / 1M tokens	27.34 s	1.03 s	19.0 tok/s	99.42%	101.4K

Missing from Ambient: MiniMax M3, Kimi K2.6.

Provider

Chutes

Offers 1 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: N/a

Review Chutes routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`chutes/int4`	Input $0.66 / 1M tokens Output $3.5 / 1M tokens	49.01 s	7.34 s	12.0 tok/s	90.48%	262.1K
GLM 5.2z-ai/glm-5.2	Not offered	—	—	—	—	—	—

Missing from Chutes: MiniMax M3, GLM 5.2.

Provider

Friendli

Offers 1 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: N/a

Review Friendli routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	Not offered	—	—	—	—	—	—
GLM 5.2z-ai/glm-5.2	`friendli`	Input $1.4 / 1M tokens Output $4.4 / 1M tokens	5.60 s	0.65 s	101.0 tok/s	99.71%	1M

Missing from Friendli: MiniMax M3, Kimi K2.6.

Provider

io.net

Offers 1 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: N/a

Review io.net routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	Not offered	—	—	—	—	—	—
GLM 5.2z-ai/glm-5.2	`io-net/fp8`	Input $1.6 / 1M tokens Output $3.5 / 1M tokens	24.86 s	2.13 s	22.0 tok/s	95.09%	262.1K

Missing from io.net: MiniMax M3, Kimi K2.6.

Provider

MiniMax

Offers 1 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: Singapore
Regions: United States

Review MiniMax routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	`minimax/fp8`	Input $0.3 / 1M tokens Output $1.2 / 1M tokens	9.82 s	1.35 s	59.0 tok/s	99.67%	524.3K
Kimi K2.6moonshotai/kimi-k2.6	Not offered	—	—	—	—	—	—
GLM 5.2z-ai/glm-5.2	Not offered	—	—	—	—	—	—

Missing from MiniMax: Kimi K2.6, GLM 5.2.

Provider

ModelRun

Offers 1 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: United States

Review ModelRun routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`modelrun/fp4`	Input $0.7 / 1M tokens Output $3.4 / 1M tokens	3.57 s	0.28 s	152.0 tok/s	100.00%	262.1K
GLM 5.2z-ai/glm-5.2	Not offered	—	—	—	—	—	—

Missing from ModelRun: MiniMax M3, GLM 5.2.

Provider

Moonshot AI

Offers 1 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: Singapore
Regions: Singapore

Review Moonshot AI routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	`moonshotai/int4`	Input $0.95 / 1M tokens Output $4 / 1M tokens	25.57 s	1.76 s	21.0 tok/s	100.00%	262.1K
GLM 5.2z-ai/glm-5.2	Not offered	—	—	—	—	—	—

Missing from Moonshot AI: MiniMax M3, GLM 5.2.

Provider

Wafer

Offers 1 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: United States
Regions: N/a

Review Wafer routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	Not offered	—	—	—	—	—	—
GLM 5.2z-ai/glm-5.2	`wafer/fp4`Alternatives: wafer/fast	Input $1.2 / 1M tokens Output $4.1 / 1M tokens	29.01 s	1.24 s	18.0 tok/s	99.73%	1M

Missing from Wafer: MiniMax M3, Kimi K2.6.

Provider

Z.ai

Offers 1 of 3 models

Speed measurements

Average 500-token response: Not comparable
Average response start: Not comparable
Average output speed: Not comparable

Reliability & location

Lowest route uptime: Not comparable
Headquarters: Singapore
Regions: Singapore

Review Z.ai routes for all selected models

Model	Selected route	Token prices	500-token response	Response starts	Output speed	Uptime	Context
MiniMax M3minimax/minimax-m3	Not offered	—	—	—	—	—	—
Kimi K2.6moonshotai/kimi-k2.6	Not offered	—	—	—	—	—	—
GLM 5.2z-ai/glm-5.2	`z-ai/fp8`	Input $1.4 / 1M tokens Output $4.4 / 1M tokens	14.82 s	3.71 s	45.0 tok/s	99.71%	1M

Missing from Z.ai: MiniMax M3, Kimi K2.6.

Only providers serving every selected model can receive a winner or metric rank. Each aggregate uses one equal 1,000-input/500-output-token request per model. Missing route measurements remain visible and are never estimated.

Comparison data fetched Jul 15, 2026, 11:25 PM UTC.

MiniMax M3, Kimi K2.6 & GLM 5.2 provider guide

How to choose one provider for MiniMax M3, Kimi K2.6 & GLM 5.2

6 providers currently offer the complete 3-model set: NovitaAI, Venice, AtlasCloud, Together, Parasail, DeepInfra. DeepInfra is closest to the current ideal combination of lower estimated model-set cost and shorter average response time.

Why complete model coverage comes first

A provider must expose MiniMax M3, Kimi K2.6, GLM 5.2 before it can receive a winner or rank. Providers missing part of the set remain available through the table toggle, but partial coverage is not compared against a complete provider portfolio.

How the recommendation is calculated

The comparison assumes one 1,000-input/500-output-token request to each model. Total cost is summed across the 3 selected routes, response time is averaged, and one recommendation is chosen as the provider closest to the normalized lower-cost and faster-response ideal.

Check exact routes and deployment requirements

Each provider row identifies the exact route selected for every model and lists alternative variants. Review quantization, context, regional hosting, privacy terms, rate limits, and current availability before standardizing this model set on one provider.

MiniMax M3, Kimi K2.6 & GLM 5.2 Provider Comparison

Choose 2–5 text models

Complete-provider winners

How to choose one provider for MiniMax M3, Kimi K2.6 & GLM 5.2

Why complete model coverage comes first

How the recommendation is calculated

Check exact routes and deployment requirements