Meta · Open weights
Llama 4 Maverick
Llama 4 Maverick, a 17 billion active parameter model with 128 experts, is the best multimodal model in its class, beating GPT-4o and Gemini 2.0 Flash across a broad range of widely reported benchmarks, while achieving comparable results to the new DeepSeek v3 on reasoning and coding—at less than half the active parameters. Llama 4 Maverick offers a best-in-class performance to cost ratio with an experimental chat version scoring ELO of 1417 on LMArena.
- Tools
- Vision
- Structured output
- Input
- Text, Images
- Output
- Text
- Discovery tier
- Open weights
Pricing
USD per 1M tokens- Input
- $0.19
- Output
- $0.65
- Cache read
- $0.05
Reference prices as published by the model provider. API example
Specifications
- Context window
- 1M
- Max output
- 65.5K
- Average output (7 days)
- 114 tokens
- Longer than 11% of 465 models
Side by side
Llama 4 Maverick compared
Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.
| Llama 4 MaverickMetaThis model | Muse Spark 1.3 ContributorMetaRemove | |
|---|---|---|
| Provider | Meta | Meta |
| Discovery tier | Open weights | Efficient |
| Added | Sep 5, 2025 | Sep 2, 2026 |
| Limits | ||
| Context window | 1.05M | 1.05M |
| Max output | 65.54K | — |
| Average output | 114 tokens | 1.2K tokens |
| Pricing, USD per 1M tokens | ||
| Input | $0.19 | $0.10Best |
| Output | $0.65 | $0.20Best |
| Cache read | $0.05 | $0.002Best |
| Performance | ||
| Throughput | — | 145.4 tok/s |
| Time to first token | — | 3.9 s |
| Request success (24h) | 99.6%Best | 99.4% |
| Benchmark indexes | ||
| Intelligence Index | 9.3 | — |
| Coding Index | 16.3 | — |
| Math Index | 19.3 | — |
| Capabilities | ||
| Reasoning | No | Yes |
| Tools | Yes | Yes |
| Vision | Yes | Yes |
| Structured output | Yes | Yes |
| PDF input | No | Yes |
| Audio input | No | Yes |
| Video input | No | Yes |
| Reasoning effort | — | minimal, low, medium, high, xhigh |
Public evaluations
Benchmarks
Independent evaluation results, with how each score ranks among the models in the Noumen catalog.
- Llama 4 Maverick
- Muse Spark 1.3 Contributor
Intelligence Index
- Llama 4 Maverick9.3
- Muse Spark 1.3 Contributor—
Llama 4 Maverick: Better than 18% of 326 scored models
Indexes
Coding Index
Composite of the coding evaluations
- Llama 4 Maverick: 16.3
- Muse Spark 1.3 Contributor: —
Math Index
Composite of the math evaluations
- Llama 4 Maverick: 19.3
- Muse Spark 1.3 Contributor: —
Agentic work
AutomationBench-AA
Workflow automation, penalized for guardrail violations
- Llama 4 Maverick: 0.2%
- Muse Spark 1.3 Contributor: —
AutomationBench-AA tasks completed
Workflows finished end to end without a guardrail violation
- Llama 4 Maverick: 0.0%
- Muse Spark 1.3 Contributor: —
Document reasoning
GDP.pdf
Professional reasoning over PDF documents, all-pass rate
- Llama 4 Maverick: 0.8%
- Muse Spark 1.3 Contributor: —
AA-LCRv1.1
Reasoning across long documents
- Llama 4 Maverick: 50.0%
- Muse Spark 1.3 Contributor: —
Reasoning and knowledge
Humanity’s Last Exam
Expert-level questions across academic fields
- Llama 4 Maverick: 4.9%
- Muse Spark 1.3 Contributor: —
GPQA Diamond
Graduate-level science questions
- Llama 4 Maverick: 67.1%
- Muse Spark 1.3 Contributor: —
MMLU-Pro
Multi-subject knowledge and reasoning
- Llama 4 Maverick: 80.9%
- Muse Spark 1.3 Contributor: —
IFBench
Following precise output instructions
- Llama 4 Maverick: 43.0%
- Muse Spark 1.3 Contributor: —
Coding
Terminal-Benchv4.0
Practical software engineering in a terminal
- Llama 4 Maverick: 0.0%
- Muse Spark 1.3 Contributor: —
SciCode
Scientific computing in Python
- Llama 4 Maverick: 31.7%
- Muse Spark 1.3 Contributor: —
LiveCodeBench
Recent competitive programming problems
- Llama 4 Maverick: 39.7%
- Muse Spark 1.3 Contributor: —
Math
AIME 2025
Competition mathematics, 2025 problems
- Llama 4 Maverick: 19.3%
- Muse Spark 1.3 Contributor: —
AIME
Competition mathematics
- Llama 4 Maverick: 39.0%
- Muse Spark 1.3 Contributor: —
MATH-500
Competition math problems
- Llama 4 Maverick: 88.9%
- Muse Spark 1.3 Contributor: —
Legacy evaluations
Terminal-Bench Hard
Earlier terminal task suite
- Llama 4 Maverick: 6.8%
- Muse Spark 1.3 Contributor: —
τ²-Bench Telecom
Tool use in a customer-support simulation
- Llama 4 Maverick: 17.8%
- Muse Spark 1.3 Contributor: —
AA-LCR (unversioned)
Long-context reasoning, earlier grading
- Llama 4 Maverick: 50.0%
- Muse Spark 1.3 Contributor: —
Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis
Live routing data
Performance and reliability
Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.
- Prompt cache hits
- 6.7%
- About $0.18 per 1M input tokens with caching
- Request success (24h)
- 99.6%
- Reliable
Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.
- Healthy (98% or more)
- Degraded (90–98%)
- Down (under 90%)
- No requests
Developers
Use it through the API
Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.
curl https://www.noumen.xyz/api/v1/chat/completions \
-H "Authorization: Bearer $NOUMEN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "meta-llama/llama-4-maverick",
"messages": [{ "role": "user", "content": "Explain this stack trace." }]
}'import os
from openai import OpenAI
client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
model="meta-llama/llama-4-maverick",
messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)Meta
Related models
Other Meta models, starting with the same family.