llama-4-maverick

1327 Score third-party Elo · arena.ai / LMArena ?

Llama 4 Maverick 17B Instruct (128E) is a high-capacity multimodal language model from Meta, built on a mixture-of-experts (MoE) architecture with 128 experts and 17 billion active parameters per forward...

LLM & Generative AI 1 model 1 listings snapshot Oct 14, 2026, 12:43 AM

Platforms

click a card to open the detail table

Cheapest OpenRouter no official benchmark Input$0.188 Output$0.652 per 1M tokens Fastest OpenRouter lowest p50 latency on this board 128.1 ms p50 Most free quota — largest published free allowance — free tier

Model variants

W3r9 dedup — same model + identical tags merge; prices are listed per billing form

Model	Provider	Tags	Input /1M	Output /1M
meta-llama/llama-4-maverick	OpenRouter	1M contextmultimodal	$0.188	$0.652

Capabilities

official design_arena & architecture metadata

Design Arena Elo official design_arena benchmark (OpenRouter)

3d 902 · 40.2% wr
uicomponent 901 · 40.8% wr
codecategories 882 · 35.8% wr
dataviz 882 · 38.4% wr
website 873 · 34.4% wr
gamedev 842 · 33.7% wr

Input / Output modalities modality: text+image->text

text in+out
image input only

Platform details

per provider

provider	model	Input /1M	Output /1M	Latency ?	Official $/1M	Availability	Data time
Novita	meta-llama/llama-4-maverick-17b-128e-instruct-fp8	$0.270	$0.850	—	—	—	2026-10-14 00:54:15
OpenRouter	meta-llama/llama-4-maverick	$0.188	$0.652	128.1 ms	—	100% available	2026-10-14 00:43:21
Together	meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP4	$0	$0	—	—	—	2026-10-14 00:53:13

3 rows · snapshot 2026-10-14 00:43:21 · from data/public/prices.json + measurements.json