Connect

Model Benchmarks

Early-access prices are provider rates with no ModelRelay markup. Intelligence by Artificial Analysis.

Speed

Cost

Intelligence

Anthropic

Claude Fable 5

TTFT

2.63s

TPS

115.0tps

Uptime

Intelligence

59.9

Unretouched model output

Input / 1M

$10.00

Output / 1M

$50.00

Anthropic

Claude Opus 5

TTFT

1.84s

TPS

115.8tps

Uptime

Intelligence

60.7

Unretouched model output

Input / 1M

$5.00

Output / 1M

$25.00

Anthropic

Claude Sonnet 5

TTFT

1.94s

TPS

150.4tps

Uptime

Intelligence

53.4

Unretouched model output

Input / 1M

$2.00

Output / 1M

$10.00

Google AI Studio

Gemini 3.5 Flash-Lite

TTFT

0.48s

TPS

439.3tps

Uptime

Intelligence

36.5

Unretouched model output

Input / 1M

$0.30

Output / 1M

$2.50

Google AI Studio

Gemini 3.6 Flash

TTFT

0.84s

TPS

272.5tps

Uptime

Intelligence

50.1

Unretouched model output

Input / 1M

$1.50

Output / 1M

$7.50

OpenAI

GPT-5.6 Luna

TTFT

1.11s

TPS

171.3tps

Uptime

Intelligence

51.2

Unretouched model output

Input / 1M

$0.20 ≤272K input

Output / 1M

$1.20 ≤272K input

OpenAI

GPT-5.6 Sol

TTFT

1.00s

TPS

117.7tps

Uptime

Intelligence

58.9

Unretouched model output

Input / 1M

$5.00 ≤272K input

Output / 1M

$30.00 ≤272K input

OpenAI

GPT-5.6 Terra

TTFT

0.78s

TPS

155.9tps

Uptime

Intelligence

55.0

Unretouched model output

Input / 1M

$2.00 ≤272K input

Output / 1M

$12.00 ≤272K input

xAI

Grok 4.5

TTFT

1.51s

TPS

71.8tps

Uptime

Intelligence

53.8

Unretouched model output

Input / 1M

$2.00 ≤200K input

Output / 1M

$6.00 ≤200K input

Swipe to compare all columns →


Anthropic	Claude Fable 5	2.63s — evidence: controlled · n=8	115.0tps — evidence: controlled · n=8	$10.00	$50.00	59.9
Anthropic	Claude Opus 5	1.84s — evidence: controlled · n=8	115.8tps — evidence: controlled · n=8	$5.00	$25.00	60.7
Anthropic	Claude Sonnet 5	1.94s — evidence: controlled · n=8	150.4tps — evidence: controlled · n=8	$2.00	$10.00	53.4
Google AI Studio	Gemini 3.5 Flash-Lite	0.48s — evidence: controlled · n=8	439.3tps — evidence: controlled · n=8	$0.30	$2.50	36.5
Google AI Studio	Gemini 3.6 Flash	0.84s — evidence: controlled · n=8	272.5tps — evidence: controlled · n=8	$1.50	$7.50	50.1
OpenAI	GPT-5.6 Luna	1.11s — evidence: controlled · n=8	171.3tps — evidence: controlled · n=8	$0.20 ≤272K input	$1.20 ≤272K input	51.2
OpenAI	GPT-5.6 Sol	1.00s — evidence: controlled · n=8	117.7tps — evidence: controlled · n=8	$5.00 ≤272K input	$30.00 ≤272K input	58.9
OpenAI	GPT-5.6 Terra	0.78s — evidence: controlled · n=8	155.9tps — evidence: controlled · n=8	$2.00 ≤272K input	$12.00 ≤272K input	55.0
xAI	Grok 4.5	1.51s — evidence: controlled · n=6	71.8tps — evidence: controlled · n=6	$2.00 ≤200K input	$6.00 ≤200K input	53.8

TTFT is time to first visible answer token; TPS is visible answer tokens per second after that token. Both use a fixed prompt at each model's minimum reasoning. Speed combines both phases into time to a 500-token answer (TTFT + 500/TPS); less time ranks higher, and neither column changes meaning. Intelligence is the Artificial Analysis Intelligence Index at each model's highest scored variant.

One integration

One API, with drop-in compatibility.

Use ModelRelay's Responses API directly, or point an existing OpenAI or Anthropic SDK at the compatible endpoints.

Call any model

ModelRelay · POST /api/v1/responses Drop-in · OpenAI /v1/responses · Anthropic /v1/messages

curl https://api.modelrelay.ai/api/v1/responses \
-H "Authorization: Bearer $MODELRELAY_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"claude-sonnet-5","input":"Hello"}'

Change the model. Keep everything else.

Build across models without rebuilding your stack.

ModelRelay is invite-only while we onboard early teams.

Request access