Skip to content
ProviderBench home

Same-model provider benchmark

DeepInfra vs MARA: LLM provider comparison

Compare DeepInfra and MARA on 4 exact shared text models. ProviderBench keeps speed, price, and catalog coverage separate so naturally faster model catalogs cannot distort the result.

DeepInfra

87 indexed models

Headquarters
United States
Server regions
N/a
Model types
67 text, 15 embeddings, 5 speech

MARA

4 indexed models

Headquarters
United States
Server regions
United States
Model types
4 text

At a glance

Metric winners

There is no overall score. Each winner answers one specific question using only directly comparable data.

Fastest on shared models

MARA

2.82× typical advantage

4 exact models with complete recent speed data

Lowest token cost

DeepInfra

$0.387 / 1M

4 exact models using a 1K-input/500-output mix

Most models available

DeepInfra

87 models

Complete catalog coverage across all indexed modalities

Shared-model benchmark summary

MetricDeepInfraMARA
Typical 500-token response14.49 s6.52 s
Typical response start0.80 s1.42 s
Typical output speed37.0 tok/s147.0 tok/s
Blended token price$0.387 / 1M$0.5692 / 1M
Typical route uptime98.77%76.79%

Visual comparison

Price and performance charts

DeepInfraMARA
500-token response by shared model

Estimated seconds using recent median response-start and output-speed data. Lower is better.

MARA has the lower typical same-model response ratio across 4 measured models.

Blended token price by shared model

USD per 1 million tokens using a 1,000-input/500-output mix. Lower is better.

DeepInfra has the lower typical price ratio across 4 priced shared models.

Shared text models

4 exact models · newest first

ModelDeepInfraMARA
MiniMax M2.7Winner · MARA

minimax/minimax-m2.7

DeepInfra
Blended price
$0.5 / 1M
500-token response
16.88 s
Response starts
0.75 s
Output speed
31.0 tok/s
Route uptime
98.77%
Context
196.6K
Route
deepinfra/fp8
MARA
Blended price
$0.48 / 1M
500-token response
3.88 s
Response starts
1.18 s
Output speed
185.0 tok/s
Route uptime
99.84%
Context
196.6K
Route
mara
MiniMax M2.5Winner · MARA

minimax/minimax-m2.5

DeepInfra
Blended price
$0.4833 / 1M
500-token response
11.21 s
Response starts
1.01 s
Output speed
49.0 tok/s
Route uptime
100.00%
Context
196.6K
Route
deepinfra/fp8
MARA
Blended price
$0.48 / 1M
500-token response
3.53 s
Response starts
1.07 s
Output speed
203.0 tok/s
Route uptime
N/a
Context
196.6K
Route
mara
DeepSeek V3.1Winner · MARA

deepseek/deepseek-chat-v3.1

DeepInfra
Blended price
$0.4833 / 1M
500-token response
42.51 s
Response starts
0.85 s
Output speed
12.0 tok/s
Route uptime
99.78%
Context
163.8K
Route
deepinfra/fp4
MARA
Blended price
$0.9667 / 1M
500-token response
17.30 s
Response starts
1.67 s
Output speed
32.0 tok/s
Route uptime
58.22%
Context
131.1K
Route
mara
gpt-oss-120bWinner · DeepInfra

openai/gpt-oss-120b

DeepInfra
Blended price
$0.0813 / 1M
500-token response
12.10 s
Response starts
0.47 s
Output speed
43.0 tok/s
Route uptime
94.64%
Context
131.1K
Route
deepinfra/bf16
MARA
Blended price
$0.35 / 1M
500-token response
9.15 s
Response starts
4.57 s
Output speed
109.0 tok/s
Route uptime
76.79%
Context
131.1K
Route
mara

A per-model winner combines blended price and estimated 500-token response time with equal proportional weight. Ties and rows missing either measurement receive no badge. Comparison data calculated . Values use one deterministic route per provider and model; missing measurements remain visible as N/a.

DeepInfra vs MARA analysis

How DeepInfra and MARA compare for AI inference

DeepInfra and MARA share 4 indexed text models, including MiniMax M2.7, MiniMax M2.5, DeepSeek V3.1, gpt-oss-120b. MARA has the stronger typical response-time result on the directly measured set.

Same-model speed evidence

4 shared models currently have complete response-start and output-speed measurements on both providers. The speed comparison uses per-model ratios before taking the median, so naturally faster model catalogs do not improve the result.

Token pricing on one workload

4 shared models have complete input and output prices on both providers. Prices use the same 1,000-input/500-output-token mix and are normalized to one million tokens for readability.

Catalog and deployment differences

DeepInfra has 87 indexed models and lists no published server regions; MARA has 4 models and lists 1 region. Verify data residency, privacy terms, limits, and production latency directly before choosing.

Related same-model benchmarks

Compare with other providers

Explore qualified alternatives with the most shared measured models. Recommendations include comparisons for both DeepInfra and MARA.

DeepInfra vs NovitaAI

45 shared model benchmarks

MARA vs AtlasCloud

3 shared model benchmarks

DeepInfra vs Venice

26 shared model benchmarks

MARA vs SambaNova

3 shared model benchmarks

DeepInfra vs SiliconFlow

23 shared model benchmarks

MARA vs Weights & Biases

3 shared model benchmarks