Llama 4 Maverick logo

Llama 4 Maverick

Meta · Released Apr 2025

Conditional

The leading open-source model for teams that need self-hosting, data sovereignty, or want to avoid API vendor lock-in.

Is it right for you?

Good for

  • Self-hosted inference
  • Custom fine-tuning for domain-specific tasks
  • Data sovereignty compliance
  • Reducing API costs at scale

Not good for

  • Tasks requiring the highest quality reasoning
  • Teams without GPU infrastructure
  • Quick prototyping where API access is faster

How it performs by task

Self-hosted inference

Excellent

Best open-weight model for production self-hosting with strong performance per compute dollar.

Fine-tuning

Excellent

Open weights enable full fine-tuning for domain-specific tasks with strong base performance.

Code generation

Good

Competent at standard code generation but behind Claude and GPT-4o on complex architectural tasks.

Data extraction

Good

Reliable for structured extraction with proper prompting, though schema adherence is less consistent.

Multilingual tasks

Very Good

Strong multilingual support across major languages with competitive fluency.

Pricing

Input

Free (self-hosted) or $0.20 / 1M tokens (hosted)

Output

Free (self-hosted) or $0.60 / 1M tokens (hosted)

Context

10M tokens

View full pricing

Benchmarks

BenchmarkScoreSource
MMLU-Pro73.7% Source
HumanEval78.6% Source
LiveBench Reasoning76.6% Source

No verdict changes yet

The clock starts day one — changes land here as our verdict evolves.

Sources

Verification log

  • Pricing— No changes

    Automated agent

  • Pricing— No changes

    Automated agent

  • Pricing— Needs attention

    Automated agent

    Source unavailable: ai.meta.com/llama returned HTTP 400. Stored pricing (Free self-hosted / $0.20/$0.60 hosted) retained.

  • Pricing— No changes

    Automated agent

    Pricing unchanged: Free (self-hosted) or ~$0.20/$0.60 per 1M tokens hosted. Open-weight model; hosted pricing varies by provider.

  • Pricing— Updated

    Automated agent

    Context window updated: stored 1M tokens → 10M tokens. Meta officially lists 10M-token context for Llama 4 Maverick.

  • Pricing— No changes

    Automated agent

  • Pricing— No changes

    Imported at launch

  • Profile— No changes

    Imported at launch

How we evaluate